Machine Learning-Driven Advancements in Electric Motorcycles: A Systematic Review of Electric Motors, Energy Storage, Charging Technologies, and Electronic Components

Abstract

The integration of artificial intelligence (AI) and machine learning (ML) technologies is rapidly transforming the design, operation, and optimization of electric motorcycles. This review analyzes research published between 2015 and 2024, focusing on how ML algorithms enhance performance, energy efficiency, diagnostics, and charging strategies across four key domains: electric motors, energy storage, charging systems, and electronic components. The review highlights state-of-the-art solutions such as torque and range prediction using LSTM/GRU models, predictive maintenance via CNNs and autoencoders, energy flow control in hybrid battery–supercapacitor systems using reinforcement learning, and federated learning for privacy-preserving embedded applications. Comparative insights reveal quantifiable performance gains over traditional methods, while integrated frameworks are proposed for linking ML diagnostics, Vehicle-to-Grid (V2G) functionalities, and renewable energy integration. The paper concludes with targeted recommendations for future research, including lightweight edge-deployable models, Explainable AI for safety-critical applications, and the fusion of intelligent charging with eco-design principles, aiming to enable intelligent, sustainable, and high-performance electric motorcycle systems.

1. Introduction

In recent years, there has been a dynamic increase in interest in electric vehicles (EVs), primarily driven by the depletion of fossil fuel resources and growing environmental awareness related to the reduction in air pollution emissions [1,2,3,4,5,6]. Particular attention from researchers and manufacturers has been directed toward electric motorcycles, which offer an alternative to conventional internal combustion engine vehicles due to their higher energy efficiency, quiet operation, and the ability to recover energy during braking [7,8,9].

A significant technological challenge lies in the development of efficient electric motors and appropriate control strategies. The most commonly used technologies include brushless DC (BLDC) motors and switched reluctance (SRM) motors, which differ in efficiency, weight, and the cost of materials used in their construction [10,11,12]. Permanent magnet synchronous motors (PMSMs) are also widely adopted due to their high efficiency and compatibility with field-oriented control (FOC) methods [13,14,15,16].

In parallel with hardware advancements, the last decade has seen a rapid expansion in the use of artificial intelligence (AI) and machine learning (ML) methods for modeling, control, diagnostics, and energy management in electric motorcycles. These algorithms have enabled more accurate representation of nonlinear drive characteristics, real-time optimization of control parameters, predictive maintenance of motors and battery systems, and adaptive charging management—often surpassing the capabilities of conventional approaches [17,18,19,20,21,22,23,24].

Energy storage remains a crucial issue. Although lithium-ion batteries are widely used, they are characterized by limited lifespan due to degradation caused by temperature and charging cycles [25,26,27,28,29]. Hybrid solutions are being developed, such as systems combining high-energy-density batteries with supercapacitors, which enable efficient management of peak power demands and extend battery life [17,18,30]. In this context, ML-supported Battery Management Systems (BMSs) are increasingly important, enabling accurate state-of-charge (SoC) and state-of-health (SoH) estimation, as well as adaptive control to extend service life [31,32,33,34].

Battery charging for electric motorcycles involves the development of both traditional wired and innovative wireless technologies. Vehicle-to-Grid (V2G) systems are gaining importance, enabling power grid stabilization and cost optimization during operation [25,35,36,37]. Integration with renewable energy sources, such as photovoltaic and wind installations, further contributes to the reduction of CO₂ emissions and the overall cost of battery charging [38,39,40,41,42,43]. Predictive and adaptive charging strategies, supported by ML algorithms, have been demonstrated to improve efficiency and flexibility in grid-connected and off-grid scenarios [44,45,46,47,48,49].

Another important aspect is the development of electric motorcycle components, including traction inverters, which must be modular, reliable, and energy-efficient to meet diverse technical requirements [19,50,51,52]. Recent research increasingly combines modern power electronics, such as silicon carbide (SiC)-based converters, with ML-based diagnostic systems to enable early fault detection and self-adaptive performance optimization [53,54,55,56,57,58,59,60]. Ensuring high comfort and safety for users is equally essential, which can be achieved through optimized suspension design, technical diagnostics, and effective energy management strategies [44,45,53,61].

In addition to advancements in electric motors, energy storage, charging technologies, and electronic components, recent studies have also focused on dynamic control challenges and socio-technical innovations specific to electric motorcycles. For example, slip ratio control algorithms that incorporate camber angle have been shown to improve posture stabilization and steering operability in two-wheel drive configurations, while sensorless, open-loop traction strategies in all-wheel drive systems can enhance stability under low-adhesion conditions. Structural optimization methods, such as topology-optimized chassis design, address the integration of battery packs without compromising strength or handling. Furthermore, research on the spatial optimization of charging station placement and the development of mobile application-based battery swapping platforms illustrates the growing role of digital tools and infrastructure planning. These developments indicate that the design of next-generation electric motorcycles requires a holistic approach, integrating vehicle dynamics, mechanical design, energy management, and user-oriented services into a unified and adaptive e-mobility ecosystem.

The aim of this article is to provide a comprehensive review of modern technological solutions used in electric motorcycles—from motors and energy storage systems to charging methods and supporting components—focusing on performance optimization, predictive diagnostics, and intelligent energy management. The authors analyze current challenges and highlight future development directions that may contribute to the widespread adoption of this type of vehicle.

This review places a particular emphasis on the integration of AI/ML techniques across all four domains, analyzing their role in control, diagnostics, and energy optimization of electric motorcycles.

2. Materials and Methods

This study adopts a systematic review approach focused on analyzing the literature published between 2015 and 2024. The selection of publications was conducted using the Scopus, IEEE Xplore, and Web of Science databases, based on keyword queries such as Electric Motorcycles, Electric Motors, Energy Storage, Battery Charging, Electrical Components, Energy Efficiency, and Sustainability. A total of 101 articles were included for analysis, meeting the following inclusion criteria: English language, topic relevance (electric motorcycles), and full-text availability. Review articles and popular science literature were excluded.

Document classification was carried out manually, based on abstracts and full texts, with articles categorized into four main areas: electric motors, energy storage, battery charging, and electrical components. A taxonomy-based qualitative analysis was employed to identify occurrences of terms related to energy efficiency and sustainability. The geographical distribution of publications and chronological trends in technological development were also examined.

Quantitative analysis was performed on technical parameters such as motor efficiency, battery energy density, and charging times.

Based on the conducted literature review, five key research questions were formulated:

• Which type of electric motor (BLDC, SRM, PMSM) is the most efficient in terms of energy performance and production cost for electric motorcycles?
• What energy storage technologies offer the best balance between performance, safety, and battery lifespan?
• Does the implementation of hybrid energy storage systems (e.g., battery-supercapacitor combinations) significantly improve electric motorcycle performance?
• Which charging technologies (wired, wireless, V2G) best meet user requirements and effectively support integration with the power grid and renewable energy sources?
• What are the optimal configurations of electric motorcycle components—particularly traction inverters—in terms of modularity, reliability, and energy efficiency?

2.1. Article Selection and Research Category Definition

The literature review was conducted using the Scopus database based on queries including the terms Electric Motorcycles, Electric Motors, Energy Storage, Battery Charging, Electrical Components, Energy Efficiency, and Sustainability, covering the period from 2015 to 2024. The selected articles were grouped into four main research categories based on Scopus keywords: electric motors, energy storage, battery charging, and electrical components, with additional consideration given to energy efficiency and sustainability aspects.

The assignment of publications to each category was based on Scopus keyword classification, while the categorization for the Research Methodology section was performed manually through full-text analysis of the articles.

The following query was used:

TITLE-ABS-KEY(“Electric Motorcycles”) AND PUBYEAR > 2014 AND PUBYEAR < 2025 AND (EXCLUDE (SUBJAREA,”BIOC”) OR EXCLUDE (SUBJAREA,”NEUR”) OR EXCLUDE (SUBJAREA,”MULT”) OR EXCLUDE (SUBJAREA,”AGRI”) OR EXCLUDE (SUBJAREA,”PSYC”) OR EXCLUDE (SUBJAREA,”ARTS”) OR EXCLUDE (SUBJAREA,”MEDI”) OR EXCLUDE (SUBJAREA,”CHEM”) OR EXCLUDE (SUBJAREA,”ECON”) OR EXCLUDE (SUBJAREA,”EART”) OR EXCLUDE (SUBJAREA,”CENG”) OR EXCLUDE (SUBJAREA,”BUSI”) OR EXCLUDE (SUBJAREA,”PHYS”)) AND (LIMIT-TO (LANGUAGE,”English”)) AND (EXCLUDE (DOCTYPE,”tb”)) AND (LIMIT-TO (EXACTKEYWORD,”Electric Motors”) OR LIMIT-TO (EXACTKEYWORD,”Traction Motors”) OR LIMIT-TO (EXACTKEYWORD,”Electric Traction”) OR LIMIT-TO (EXACTKEYWORD,”Reluctance Motors”) OR LIMIT-TO (EXACTKEYWORD,”Electric Drives”) OR LIMIT-TO (EXACTKEYWORD,”Energy Storage”) OR LIMIT-TO (EXACTKEYWORD,”Secondary Batteries”) OR LIMIT-TO (EXACTKEYWORD,”Lithium-ion Batteries”) OR LIMIT-TO (EXACTKEYWORD,”Supercapacitor”) OR LIMIT-TO (EXACTKEYWORD,”Battery Pack”) OR LIMIT-TO (EXACTKEYWORD,”Electric Batteries”) OR LIMIT-TO (EXACTKEYWORD,”Battery”) OR LIMIT-TO (EXACTKEYWORD,”Hybrid Energy Storage Systems”) OR LIMIT-TO (EXACTKEYWORD,”Charging (batteries)”) OR LIMIT-TO (EXACTKEYWORD,”Battery Management Systems”) OR LIMIT-TO (EXACTKEYWORD,”Regenerative Braking”) OR LIMIT-TO (EXACTKEYWORD,”Charging Station”) OR LIMIT-TO (EXACTKEYWORD,”Battery Swapping”) OR LIMIT-TO (EXACTKEYWORD,”Battery Chargers”) OR LIMIT-TO (EXACTKEYWORD,”Permanent Magnets”) OR LIMIT-TO (EXACTKEYWORD,”Electric Inverters”) OR LIMIT-TO (EXACTKEYWORD,”DC-DC Converters”) OR LIMIT-TO (EXACTKEYWORD,”Digital Storage”))

In the next stage, selection was carried out based on the results of the Scopus database query, which initially yielded 194 publications published between 2015 and 2024, written in English and classified within subject areas such as Engineering, Energy, Computer Science, Mathematics, Social Sciences, Environmental Science, Materials Science, and Decision Sciences. To narrow the thematic scope, the keywords Electric Motorcycles, Electric Motors, Energy Storage, Battery Charging, Electrical Components, Energy Efficiency, and Sustainability were applied. Non-technical fields (e.g., biochemistry, psychology, arts, medicine) were excluded during the filtering of Scopus subdisciplines.

The titles and abstracts of the publications were extracted from the Scopus database and filtered using technical and thematic criteria. The initial set of 194 publications was reduced to 101 items by applying an additional keyword filter in Scopus. This process followed steps inspired by the PRISMA guidelines, including identification, screening, and eligibility assessment.

Furthermore, the full texts of the publications were manually reviewed to confirm compliance with the adopted analytical criteria. Titles and abstracts underwent initial screening to remove duplicates as well as review and popular science articles. Subsequently, full texts were manually verified, resulting in a refined set of 101 articles meeting the required technical and thematic standards.

The selected articles were assigned to two main thematic categories based on content analysis: (1) E-mobility Technologies—encompassing electric motors, energy storage, battery charging, and electrical components; (2) Engineering and Optimization—focusing on energy efficiency, sustainability, design, and methodological analysis, including topology, simulation, and the finite element method.

A geographical analysis of author affiliations was performed for countries including Canada, India, Indonesia, Italy, Japan, Malaysia, South Korea, Spain, Thailand, the United Kingdom, and the United States, enabling an assessment of the global distribution of research.

A visualization of the literature selection stages, applied filtering criteria, and the main classification directions is presented in Figure 1, which also illustrates the thematic distribution of the analyzed publications. The final set of 101 articles was considered representative for the subsequent review analysis.

2.2. Classification Criteria

As part of the literature review, 194 documents were identified in the Scopus database using a query combining the terms Electric Motorcycles, Electric Motors, Energy Storage, Battery Charging, Electrical Components, Energy Efficiency, and Sustainability. The results were limited to publications from 2015 to 2024, written in English, and classified within the subject areas of Engineering, Energy, Computer Science, Mathematics, Social Sciences, Environmental Science, Materials Science, and Decision Sciences. The query was subsequently refined by adding keywords from the Engineering and Optimization group, including Design, MATLAB, Topology, Simulation, Finite Element Method, Energy Utilization, Performance, Energy Efficiency, Efficiency, Power, Cost Effectiveness, Greenhouse Gases, Internal Combustion Engines, Fossil Fuels, Global Warming, Gas Emissions, Costs, and Electric Utilities. This refinement reduced the dataset to 101 documents.

Further narrowing was achieved by excluding literature from non-technical domains (e.g., AGRI, PSYC, ARTS, MEDI, ECON, EART, CENG, BUSI, PHYS), thereby ensuring high topical relevance.

For the purposes of further analysis, five classification criteria were adopted: thematic group, country of author affiliation, document type, research methodology, and keyword set.

The first criterion, thematic grouping, distinguished two main categories:

• E-mobility Technologies:
  • Electric Motors
  • Energy Storage
  • Battery Charging
  • Electrical Components
• Engineering and Optimization:
  • Energy Efficiency
  • Sustainability
  • Design and Analysis

The second criterion concerned the country of author affiliation, determined based on Scopus bibliographic data and including, among others, Canada, India, Indonesia, Italy, Japan, Malaysia, South Korea, Spain, Thailand, the United Kingdom, and the United States. Remaining countries were grouped under the category “Other.”

The third criterion was document type, encompassing journal articles and conference papers.

The fourth criterion referred to research methodology, identified through content analysis and including experimental approaches, literature reviews, case studies, conceptual works, and surveys.

The fifth criterion was based on the keywords used in the publications, which were grouped into the main categories of E-mobility Technologies and Engineering and Optimization, allowing for the inclusion of diverse research approaches.

By applying these well-defined criteria, a comprehensive and comparative analysis of research on electric motorcycle technologies was made possible, forming the foundation for the conclusions drawn in this review. This approach ensured methodological consistency and transparency throughout the literature review process. The classification results are presented in tabular form in Section 3 and discussed in the context of technological and geographical trends.

2.3. Data Processing and Analysis

The bibliographic data retrieved from the Scopus database, comprising 101 scientific publications that met the established thematic, linguistic, and temporal criteria, underwent a process of standardization and preparation for analysis.

In the initial phase, manual verification of titles, abstracts, and keywords was conducted, and each publication was assigned at least one classification category based on keywords and source metadata. The categories identified included e-mobility technologies (electric motors, energy storage, battery charging, electrical components) and engineering and optimization (energy efficiency, sustainability, design and analysis).

Additionally, the type of document (journal article, conference paper), country of author affiliation (including Canada, India, Indonesia, Italy, Japan, Malaysia, South Korea, Spain, Thailand, the United Kingdom, and the United States), and research methodology (experiment, literature review, case study, conceptual paper, survey) were classified.

Data processing and visualization were carried out using Python 3.12, along with the Pandas 2.2.2, Matplotlib 3.8.4, and NumPy 1.26.4 libraries. The data were imported into a local PostgreSQL 16.2 database, in which relational tables were created to store information on authors, affiliations, keywords, publication year, and DOI identifiers. SQL queries enabled cross-sectional thematic analyses and comparisons across selected categories.

Geographical analysis enabled the identification of publication distribution by country and the most active research institutions. To monitor temporal developments, a year-by-year analysis covering the period 2015–2024 was conducted, allowing the identification of trends and research dynamics in the field of e-mobility technologies.

Network graphs generated using VOSviewer 1.6.20 were employed to visualize the co-occurrence of keywords.

Through the synergy of programming tools, database systems, and manual classification, an in-depth insight into the structure and development directions of research on electric motorcycles and related technologies was obtained.

2.4. Review Protocol and Publication Quality

To ensure transparency and reproducibility of the procedure, the selection and inclusion process was organized in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) protocol. The diagram illustrates the four fundamental stages of the process:

• Identification—A total of 194 records were identified in the Scopus database based on search criteria related to titles, abstracts, and keywords associated with electric motorcycles.
• Screening—After removing duplicates (n = 0), a preliminary screening of titles and abstracts was conducted (n = 194), resulting in the exclusion of 93 records that did not meet the study’s inclusion criteria.
• Eligibility—101 full-text articles were assessed for substantive relevance to the defined research scope. No further exclusions were necessary at this stage.
• Included—All 101 articles were ultimately included in the qualitative synthesis.

The selection process is illustrated in Figure 2, which provides a structured overview of the flow from initial identification to final inclusion, helping to clarify both the rigor and transparency of the review methodology.

The second part of the analysis focused on the thematic classification of publication content based on keywords and abstracts. VOSviewer was used to identify dominant concepts and term co-occurrence networks, which allowed for the extraction of key research topics. The analysis revealed two main thematic axes: e-mobility technologies and energy efficiency and optimization. Within the first axis, categories related to electric motors, energy storage, battery charging, and electrical components were identified. The second axis included topics such as energy efficiency, sustainability, system design, and performance analysis.

Geographical classification was based on the affiliations of the authors of each publication. The most frequently represented countries included Indonesia, Japan, the United Kingdom, the United States, Canada, Italy, Thailand, India, Malaysia, South Korea, and Spain. Publications from countries with only a single representation were grouped under the category “Other.” This distinction enabled the identification of geographical centers with the highest research activity in the field of electric motorcycles.

Documents were also classified by publication type: journal articles and conference papers.

The research methodology was determined based on abstract content and full-text analysis, considering five categories: experimental studies, literature reviews, case studies, conceptual papers, and survey-based research.

For the purposes of this classification, the data were manually verified by two independent reviewers, who assigned categories based on predefined criteria. In cases of disagreement, consensus was reached through discussion. This approach ensured consistency in classification and enabled a multidimensional analysis of the literature, encompassing technical, geographical, and methodological aspects.

The categorization described was based not only on the presence of specific terms but also on their frequency and position within the keyword co-occurrence network, as visualized in the keyword density map presented in Figure 3. This analysis facilitated the identification of the most frequently addressed research areas as well as existing knowledge gaps requiring further investigation.

The density analysis generated in VOSviewer presents the distribution and frequency of key terms within the corpus of 101 publications on electric motorcycles. The map uses a color scale ranging from blue (low frequency), through green (moderate frequency), to yellow (high frequency), enabling rapid identification of the most significant concepts.

At the center of the diagram is the term “electric motorcycles”, surrounded by a yellow area indicating the highest term density, which reflects its dominant role in the analyzed dataset. Closely surrounding it are “electric motors” and “lithium-ion batteries”, which also appear with high intensity (green-to-yellow areas), highlighting strong research interest in electric drives and energy storage systems.

In the upper-left section of the map, a thematic cluster is distinguished that relates to drive system design, including terms such as “traction motors”, “reluctance motors”, “permanent magnets”, and “finite element method”. This cluster reflects intensive exploration of motor types, and the methods used to analyze their properties.

The lower section of the diagram features terms like “energy utilization” and “electric drives”, indicating research focused on optimizing system efficiency.

The right-hand side of the map contains environmentally and logistically oriented terms such as “charging (batteries)”, “greenhouse gases”, “gas emissions”, and “cost effectiveness”. These appear in green shades, suggesting a moderate frequency of occurrence and growing interest in the environmental and economic impacts of electromobility.

Smaller, blue areas of the map include simulation tools (“simulation”, “MATLAB”), electrical components (“electric inverters”), and battery management topics (“battery management systems”, “battery pack”), underlining the multidisciplinary nature of the research.

The entire map illustrates the complex structure of electric motorcycle research, where clear concentration around core topics related to hardware, energy systems, and environmental impact enables the identification of primary development directions and research gaps.

Interpreting this network reveals not only the key research domains but also potential thematic gaps—for example, the relatively low number of connections between the “battery management systems” cluster and “blockchain” may indicate the need for more in-depth exploration of distributed ledger technologies in the context of energy management in electric vehicles.

Thus, Figure 4 provides a visual perspective on the structure and complexity of the analyzed literature, supporting a quick understanding of inter-topic relationships and highlighting areas for further investigation.

The network diagram generated in VOSviewer illustrates the co-occurrence relationships among 28 key terms extracted from the corpus of 101 publications. The size of each node corresponds to the frequency of the term in the analyzed dataset, while the thickness and length of the links represent the strength and frequency of co-occurrence between term pairs. The colors of nodes and edges indicate thematic clusters automatically detected by the VOSviewer algorithm.

Blue cluster (center, largest node “electric motorcycles”): The central node of the network is “electric motorcycles”, reflecting its dominant role in the literature. It is directly connected to terms such as “lithium-ion batteries”, “performance”, and “battery management systems”, highlighting a close relationship between research on electric drivetrains, energy storage, and battery management systems.

Purple cluster (bottom of the diagram): Nodes such as “electric motors”, “energy utilization”, “electric batteries”, “simulation”, and “MATLAB” form a distinct research area focused on modeling, numerical simulations, and the analysis of energy-efficient drive systems. The thick link between “electric motors” and “energy utilization” indicates a frequent co-occurrence of these topics in the literature.

Red cluster (upper left area): The term “traction motors” serves as the central node in a cluster centered on technical discussions related to traction motors, including “reluctance motors” and “permanent magnets”. The presence of “finite element method” and “topology” in the same cluster suggests the widespread use of FEM and topology optimization methods in motor design research.

Green cluster (right area): This segment of the network focuses on environmental and operational topics, including “charging (batteries)”, “greenhouse gases”, “gas emissions”, “cost effectiveness”, “battery swapping”, and “electric utilities”. The links between these nodes indicate an interdisciplinary approach that integrates charging technology with environmental impact and cost analysis.

Yellow cluster (upper right section): A smaller but distinct cluster surrounding “energy efficiency”, “battery pack”, and “supercapacitor” reflects growing interest in research aimed at improving energy efficiency and designing advanced storage systems, including hybrid battery–supercapacitor architectures.

This network structure enables an understanding of how core technical topics (drivetrains, batteries, simulations) connect with environmental and economic concerns (emissions, costs, energy services) and reveals both strong and weaker thematic interrelations. The network diagram (Figure 4) facilitates the rapid identification of central research areas and their interconnections.

The bibliographic references marked as [4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,28,29,30,31,32,33,34,36,37,39,40,41,42,43,44,45,46,47,48,49,51,52,53,54,55,56,57,58,59,60,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112] were included in the final dataset for the systematic review, allowing readers to clearly distinguish the core set of analyzed studies from supporting or contextual literature.

3. State of the Art: Review from 2015 to 2024

This section presents an overview of the most significant research advancements in the field of electric motorcycles and related technologies between 2015 and 2024. The analysis includes studies focused on the design of drivetrain systems, including hybrid and fully electric powertrains utilizing PMSM, BLDC, and SRM machines, as well as advanced traction inverters. Particular emphasis is placed on research into energy storage systems—from lithium-ion battery packs and hybrid configurations with supercapacitors, to battery capacity selection methodologies and strategies for managing temperature and state of charge, often implemented in modular BMS architectures.

Another important research stream involves modeling and simulation (MATLAB/Simulink, FEM, VOSviewer) aimed at evaluating performance, energy efficiency, and vehicle dynamics, alongside experimental studies that validate simulation outcomes.

Attention has also been drawn to the development of charging infrastructure—ranging from low-cost grid- and photovoltaic-based stations, to range-extending devices and the economic and environmental aspects of charging within the broader context of global electromobility trends.

Finally, the chapter discusses modern approaches to speed control and ride safety, including adaptive neuro-fuzzy systems, predictive control, and optimal control methods, collectively providing a comprehensive picture of research into electric motorcycles.

3.1. Application of Machine Learning Algorithms in Modeling and Control of Electric Motors

In the past decade, there has been a rapid development in the application of machine learning (ML) algorithms for modeling and controlling electric motors used in electric vehicles, particularly in electric motorcycles. These algorithms have enabled more accurate representation of the nonlinear characteristics of drives, real-time dynamic control, and increased energy efficiency and component lifespan.

Field-Oriented Control (FOC) was applied to a PMSM, demonstrating its superiority over other methods in terms of smooth operation and torque ripple reduction [13]. A modular, scalable inverter was developed in, optimized for efficiency and compatibility with various motor types [19]. Study [20] combined artificial neural networks (ANN) with a genetic algorithm (GA), resulting in significant improvements in energy efficiency and range for electric scooters.

Optimal control models were employed to enhance the driving strategy of a racing motorcycle in [64]. Predictive models such as LSTM and fuzzy logic systems were used in [17], where a hybrid energy management system (HESS) was developed to predict power distribution between the battery and supercapacitors.

Motor temperature prediction models under variable load were introduced in, which could be developed using nonlinear regression or deep neural networks (DNN) [65]. Study [18] focused on structural optimization of motorcycle frames, highlighting the potential of PSO algorithms for designing lightweight yet robust structures. Alternative drive architectures and the impact of gear selection and energy recovery mechanisms on the energy efficiency of electric motorcycles were analyzed in [8,31,51,52]. In contrast, [28,67,69,71] focused on structural optimization—from cooling systems to frame geometries—using ML algorithms for simulation and design optimization.

The use of real-world data also played a significant role. In [68], motor operating parameters were modeled using LabVIEW, suggesting their potential use as input for predictive ML models. Electric motorcycle riding patterns were classified in based on GPS and accelerometer data, supporting dynamic optimization of drive operations using algorithms such as SVM or XGBoost [73].

Hybrid models (e.g., ANFIS) were proposed in [21], where features of fuzzy systems and neural networks were combined for precise BLDC motor control. Study [14] analyzed telemetry data from races, indicating the potential for integrating video data with ML to model power profiles. The effectiveness of classifiers in predicting motor faults based on vibration signals using CNNs and autoencoders was demonstrated in [6,77].

The use of XGBoost and Random Forest algorithms to assess energy efficiency in motorcycles was presented in [9,79], while LSTM and GRU models for torque prediction were described in [22,82]. FOC and DTC were compared as drive control strategies in [15], emphasizing the possibility of hybridization supported by learning algorithms.

Control systems based on adaptive ML algorithms that adjust to changing road conditions were analyzed in [23,42,52]. Reinforcement Learning was applied for real-time motor control in [24], whereas [86] presented the application of federated learning to update ML models without transmitting raw data to the cloud. The concept of distributed learning of predictive models in environments with limited connectivity was developed in [43,87,88]. In [21,91,92,93] the importance of data security and privacy was emphasized in locally trained models, including the use of homomorphic encryption and differential privacy.

Further studies such as [37,89,90] utilized sensor data topologies from IPM and SRM drives for dynamic predictive model training. The impact of optimizing input data (e.g., driving trajectory, mass, battery state) on the accuracy of ML models was analyzed in [91].

The use of classifiers to identify load types and recognize terrain conditions was described in [18,95,96], while ensemble learning methods to improve model accuracy and robustness were presented in [98,99]. Additional ensemble learning techniques and bagging methods were proposed in [97,100,101,102] for load and operating condition classification, as well as improved robustness of models to noisy or incomplete data. Finally, in [12,60,103,104,112], signal transformation techniques (e.g., FFT, DWT) were applied to improve the quality of input data used in predictive ML models.

In addition to the previously described ML applications in controlling PMSMs, BLDC motors, and SRM motors, significant contributions address dynamic challenges specific to electric motorcycles. For instance, a slip ratio control algorithm explicitly considering the camber angle was developed for a two-wheel drive electric motorcycle, improving posture stabilization and steering operability by adapting control strategies to the unique dynamics of single-track vehicles [31,113]. A similar approach, but implemented in an all-wheel drive architecture, was presented in a study proposing a sensorless, open-loop traction strategy. Simulation results demonstrated that such an AWD configuration can significantly enhance stability and performance, particularly under low-adhesion conditions. Recent work has also explored the integration of advanced sensing solutions to monitor dynamic and electrical parameters in real time [114], as well as optimizing control electronics to reduce switching losses in motorcycle drive controllers [77,115]. These examples illustrate how ML-driven and advanced control methods are increasingly adapted to the specific dynamic requirements of electric motorcycles, expanding current research on drive architecture optimization.

In summary, research from 2015 to 2024 demonstrated a significant expansion in the use of ML in controlling electric motorcycle drives—from traditional regression algorithms to deep networks, reinforcement learning, and federated learning. Their effectiveness depended on data quality, feature selection, and implementation capabilities in embedded systems.

Table 1. Types of Machine Learning Algorithms and Their Applications in the Prediction of E-mobility Technologies.

Type of ML Algorithm	Application	Example Publications
Linear/Nonlinear Regression	Prediction of energy consumption, torque, motor temperature	[9,13,65]
Neural Networks (ANN, DNN)	Drive behavior modeling, parameter estimation, adaptive control	[20,21,22,89]
Genetic Algorithms (GA)	Optimization of drive configuration and control parameters	[20,39,91]
XGBoost, Random Forest	Efficiency prediction, driving style classification, operational data analysis	[9,73,79,95]
SVM	Driving pattern classification, road condition detection	[18,23,73]
LSTM, GRU	Range, torque, SoC prediction, time series control	[22,64,82,86]
Fuzzy Logic/ANFIS	Fuzzy drive control, adaptation to operating conditions	[17,21,23]
CNN	Classification of acoustic, image, vibration data, sensor data analysis	[14,37,77]
Autoencoders	Fault and anomaly detection in motors	[6,98]
Reinforcement Learning	Real-time drive control learning	[24,103]
Federated Learning	Remote ML model training without data centralization	[12,86]
Ensemble Learning	Improving classifier accuracy and robustness	[60,99]

presents a summary of the types of machine learning algorithms and their applications in E-mobility technologies.

The application of machine learning (ML) algorithms in modeling and controlling electric motors in electric vehicles has become one of the key tools supporting the development of electromobility. ML has enabled precise mapping of the nonlinear characteristics of drive systems, real-time adaptive control parameters depending on driving conditions, and predictive management of motor operating parameters. Particularly effective results have been demonstrated with the implementation of neural networks, genetic algorithms, LSTM, CNN, SVM, and hybrid methods based on Fuzzy Logic and ANFIS.

The research addressed both structural and operational aspects—from drive system design optimization to torque and speed control, and real-time anomaly detection. ML algorithms have also enabled the integration of data from sensors, GPS systems, video, and accelerometers, facilitating adaptive and autonomous control of PMSM, BLDC, SRM, and other motor types. Additionally, ML has shown potential for deployment in embedded systems, opening paths for commercial smart drive solutions.

An illustrative example of combining structural optimization with intelligent control is presented in the design and optimization of an external-rotor switched reluctance motor (SRM) for an electric scooter, where Finite Element Analysis (FEA) and Particle Swarm Optimization (PSO) were applied to increase the average torque and reduce copper losses [91,116]. Such optimized motor configurations can be further integrated with predictive ML models, enabling adaptive adjustment of operating parameters in real time based on load, temperature, and riding conditions [117]. This approach reflects the broader trend of merging mechanical design optimization with data-driven control strategies to simultaneously improve drivetrain efficiency and durability.

In summary, machine learning has significantly enhanced the precision, responsiveness, and energy efficiency of electric drive systems. Further development should focus on improving model robustness, reducing computational requirements, and integrating with comprehensive vehicle energy and safety management systems.

3.2. Algorithms for Control and Prediction in Energy Management Systems of Electric Vehicles and Grid Integration

With the increasing demand for intelligent and flexible power systems in electromobility, machine learning (ML) algorithms have begun to play a vital role in energy management systems (EMS) and the integration of electric vehicles with power grid infrastructure. Their application has enabled energy demand forecasting, charging schedule optimization, integration with renewable energy sources (RES), and minimization of operating costs and CO₂ emissions.

Study [63] analyzed the process of energy recovery during vehicle braking, where sensor data could be used in ML models to predict the recovered power under various driving conditions. In [10], PV system selection for EVs was analyzed using probabilistic distributions, indicating the potential for extending the approach with ML algorithms to dynamically predict energy supply and demand.

In [7], SVM and k-NN were used to forecast grid loads resulting from mass EV charging. A similar approach was taken in [30], where urban electric vehicle fleet data was used to predict instantaneous energy demand. Study [44] proposed a cost optimization model for charging under dynamic tariffs, which could be enhanced with boosting and logistic regression for improved precision.

Network integration issues were addressed in [45,46], which demonstrated the potential of dynamically managing power based on predictions using LSTM and XGBoost models. In [32,47], the impact of EV integration with RES on microgrid stability was analyzed, suggesting the implementation of predictive ML algorithms using weather and historical data.

In [40,70,72] reinforcement learning was used for energy management strategies and adaptive neural networks to independently tailor energy management strategies based on user profiles. Federated learning models described in [41,75] enabled model training on vehicles without transmitting raw data to the cloud, enhancing privacy and system security.

Ensemble learning for battery health classification, lifespan estimation, and replacement timing was presented in [33,34,53,55]. The use of autoencoders and signal transformation to improve data quality in energy management was described in [5,56,76]. Additional studies [4,54,57,74] introduced advanced signal processing techniques from sensors—including DWT, STFT, and nonlinear filtering—which significantly improved ML input data quality for energy management and fault detection.

In subsequent research [9,11,58,59,78,79,80] the focus was on real-time prediction of battery and supercapacitor behavior. It was shown that LSTM, GRU, and CNN enabled accurate estimation of SoC, SoH, and thermal control of battery packs, contributing to increased lifespan and efficiency. Complementary studies [30,80,81,83] addressed battery state monitoring using fuzzy logic models, Kalman filters, and adaptive neural networks, improving SoH prediction accuracy under real-world conditions.

A notable example of advanced energy management in electric motorcycles is provided by the optimal control of fuel consumption in a hybrid motorcycle equipped with three power sources: an internal combustion engine, an electric motor, and an integrated starter generator (ISG). The study compared two energy management strategies—Rule-Based Control (RBC) and the Equivalent Consumption Minimization Strategy (ECMS)—and quantified measurable reductions in fuel consumption [80]. These findings demonstrate that, even in complex multi-source configurations, substantial efficiency gains can be achieved through the appropriate selection of control strategies, particularly when supported by data-driven prediction from real-world driving cycles. Integrating such solutions with ML approaches, such as reinforcement learning, could further enhance the system’s ability to adapt in real time to user profiles, road conditions, and charging infrastructure availability.

Publications [16,48,49,84,85] analyzed multipoint vehicle charging strategies using real-time data from chargers and vehicles. ML algorithms were used to predict peak demand, optimize power allocation, and reduce operational costs.

Finally, in [31,107,108,109,110,111], attempts were made to integrate ML with next-generation energy control systems based on distributed architectures, cloud computing, and neural networks processing sensor, weather, and tariff data in real time. Studies [105,106] further explored control in modern EMSs, describing dynamic power allocation and energy quality control using agent-based decision systems and peer-to-peer topologies.

Table 2. Types of Machine Learning Algorithms and Their Applications in the Prediction of E-Mobility Energy Management Systems.

Application Area	Objective/Function	Applied ML Algorithms	Publications
Energy Demand Prediction	Estimating instantaneous and long-term energy consumption by EVs	LSTM, XGBoost, regression, SVM	[7,30,63,78,79]
Charging Optimization	Power management, dynamic tariffs, charging schedules	Reinforcement Learning, boosting, CNN	[16,44,48,49]
Integration with RES (Renewable Energy Sources)	Matching PV/wind supply to EV demand	LSTM, hybrid models	[10,45,46,47]
Battery Diagnostics and Durability	Estimating SoH, SoC, life cycle, replacement timing	Autoencoders, ensemble, GRU	[33,34,53,58,59]
Predictive Systems in Microgrids	Load stabilization, overload and fault detection	SVM, k-NN, predictive models	[5,7,32,72]
Distributed and Private Learning	Model training without data transfer (privacy-aware EMS)	Federated Learning, edge ML	[41,75,107]
Energy Data Transformation	Improving data quality for ML (FFT, DWT, PCA)	CNN, autoencoders, PCA	[6,56,76,112]
Dynamic Control with Sensor Data	Real-time management based on GPS, IMU, thermometers, power sensors	Deep Learning, Reinforcement Learning	[40,72,85,109]

provides a comprehensive summary of the different application areas for machine learning in E-mobility energy management systems.

In the context of EMS optimization, research on the thermal management of a swappable lithium-ion battery pack for an electric motorcycle demonstrated that an air-cooling system could improve cell performance by up to 12% [29]. The integration of such cooling systems with ML-based predictive control enables proactive regulation of battery temperature and state-of-charge (SoC), extending battery lifespan and enhancing operational safety.

Another relevant consideration is the mechanical integration of the battery pack into the motorcycle chassis. A topology-optimized frame design, providing a lightweight yet stiff structure while accommodating the battery safely, addresses this challenge [96]. When combined with predictive EMS algorithms, such design solutions enable the creation of lightweight, efficient, and service-friendly electric motorcycles ready for seamless integration with microgrids and Vehicle-to-Grid (V2G) systems.

The application of machine learning algorithms in energy management systems for electric vehicles and their integration with the grid has enabled the development of intelligent, adaptive, and highly efficient solutions. By utilizing models such as LSTM, GRU, CNN, and XGBoost, it has become possible to accurately predict energy consumption, battery state-of-charge and state-of-health (SoC/SoH), as well as dynamically control the charging process. These algorithms also support energy management in grid conditions—allowing for power balancing, minimization of overloads, and adaptation to variable tariffs and weather conditions. Increasingly important are locally trained models (federated learning), which protect data privacy and enhance the security of EMSs.

Research conducted between 2015 and 2024 highlights the growing significance of deep learning and ensemble methods in the field of electric vehicle energy management. Particular attention is being paid to the development of robust, energy-efficient models deployable in embedded systems and microgrids. Future research should focus on improving model transparency (Explainable AI), resistance to interference, and adaptability to changing user conditions—elements that are essential for the widespread integration of ML in modern energy management systems within e-mobility.

Beyond purely technical optimization, machine learning can also support policy-making and infrastructure planning in the electric motorcycle domain. For example, a recent study examining the effectiveness of policies for developing the battery swapping service industry in Indonesia applied a system dynamics model to evaluate the impact of subsidies, standardization, and other regulatory measures. The findings suggest that well-designed policy instruments can accelerate technology adoption and reduce entry barriers for service providers—an approach that can be enhanced by ML-driven scenario simulations to forecast adoption rates and economic outcomes [21].

In addition, optimizing the spatial placement of electric motorcycle charging stations has been demonstrated in a case study from Surakarta, Indonesia, where centrality index and scalogram calculations were used to identify station locations that maximize accessibility and public benefit [118]. When combined with ML-based predictive models incorporating traffic density, population distribution, and charging patterns, such spatial analyses can significantly improve the efficiency and user experience of charging networks [75].

Furthermore, the concept of Tukerin, a mobile application-based marketplace for electric motorcycle battery swapping, illustrates how digital platforms can address key adoption barriers such as range anxiety and long charging times [119,120]. By integrating this type of platform with demand prediction and energy management systems powered by ML, it becomes possible to optimize battery inventory, reduce wait times, and coordinate with Vehicle-to-Grid (V2G) operations [6].

Incorporating such socio-technical innovations into the architecture of energy management systems (EMS) for electric motorcycles bridges technical optimization (charging strategies, demand prediction) with infrastructure and policy considerations, creating a holistic and adaptive ecosystem for sustainable e-mobility growth.

3.3. ML Algorithms in Control and Prediction Within Electric Vehicle Energy Management Systems and Their Integration with the Grid

With the development of electromobility and the increasing demand for intelligent, self-adaptive, and efficient control systems, energy management, and optimization of charging infrastructure, machine learning (ML) algorithms are becoming the foundation of modern engineering solutions. Future ML applications in this domain will focus on deeper integration with distributed technologies, improved model interpretability, and adaptation to variable operational and environmental conditions.

One of the key areas of development is federated learning, which enables training ML models without transferring raw data to the cloud. This allows user privacy to be preserved while still improving predictive models—an essential aspect for deploying ML in network-connected vehicle systems [41,75,107]. Such models can be dynamically updated at the level of an individual vehicle or fleet, increasing local autonomy and system resilience.

Another significant trend is the development of energy-efficient algorithms optimized for edge computing. ML models implemented directly in onboard controllers of motorcycles and other EVs must be characterized by low power consumption and high resilience to disturbances, especially in rough terrain or urban traffic overloads [66,67,68]. For this purpose, lightweight neural networks, compressed deep models, and self-adaptive systems capable of real-time operation are being developed.

Reinforcement learning is also becoming an important direction, enabling control and energy management systems to learn optimal strategies based on experience and environmental feedback [43,70,72,109]. For example, such models can be used for continuous optimization of power distribution between the battery and supercapacitors in HESSs, taking into account driving style, predicted road conditions, and charging infrastructure availability.

As the number of electric vehicles grows, intelligent charging management and operating cost optimization are also advancing. Algorithms such as XGBoost, LSTM, and CNN allow highly accurate forecasting of energy demand—even under variable tariffs, power limits, and irregular usage patterns [39,79,81]. Integration with renewable energy sources (RES), microgrids, and dynamic charging stations requires not only prediction but also real-time response, which can be achieved through hybrid ML models.

Predictive maintenance systems are also receiving increasing attention, enabling early detection of irregularities in motors, batteries, and electronic components. The use of autoencoders, recurrent neural networks, and ensemble learning makes it possible to identify subtle changes in measurement signals before they result in failure [41,63,65,71,74]. These models support maintenance planning, improve safety, and reduce unplanned downtimes.

A vital area of development is data fusion from multiple sources—onboard sensors, GPS, weather data, energy tariffs, and contextual user information. The application of deep neural networks (DNN) and dimensionality reduction methods allows for building complex yet efficient predictive models that are robust to disturbances and measurement imperfections [6,71,112].

Explainable AI (XAI) also plays a key role—referring to algorithms whose operations can be understood by end users or system operators. Transparency of ML models will be critical for their certification, deployment in safety-critical environments, and for building trust in autonomous technologies. Transparent models will find use in energy safety management systems and automated decision-support systems [40,52,54,82].

In the coming years, ML is expected to be increasingly integrated with cloud-based architectures, IoT, and distributed systems. These solutions will form the foundation of fully intelligent EV fleet management systems, integrating energy management, technical maintenance, and adaptive drive control within one cohesive environment.

3.4. Summary of the Analysis in Relation to Key Research Questions

In the context of electric motor efficiency used in electric motorcycles, the most optimal solution in terms of energy performance remains the Permanent Magnet Synchronous Motor (PMSM). These motors are characterized by high power density and efficiency, making them an attractive choice for premium and sport segments. Their advantages also include low noise levels and favorable control characteristics under variable operating conditions [14,67,68,73]. While Brushless DC motors (BLDC) are slightly less efficient, they are simpler in design, cheaper to manufacture, and easier to maintain, making them ideal for more economical vehicle models [13,15]. Switched Reluctance Motors (SRM), on the other hand, are notable for their high durability and resistance to external factors, although their development is limited by noise emission challenges and complex control characteristics [23,91].

In the field of energy storage technologies, lithium-ion cells remain the dominant solution due to their favorable capacity-to-weight ratio, long lifecycle, and good thermal performance [64,65,66]. Safer alternatives include LFP cells and solid-state electrolyte systems, which offer higher thermal stability and lower fire risk, although often at the expense of energy density [39,45,51]. Increasing attention is being given to advanced Battery Management Systems (BMS) supported by ML algorithms, which enable predictive control of battery parameters and monitoring of their condition [32,46].

The use of Hybrid Energy Storage Systems (HESS), integrating batteries with supercapacitors, significantly enhances the performance of electric motorcycles, especially in the context of regenerative braking and the delivery of high instantaneous power during acceleration. Such systems increase the lifespan of the main battery pack by reducing peak loads [31,47,70]. Combined with ML algorithms, they enable dynamic energy flow management by adapting to riding style, road conditions, and the vehicle’s energy demand [40,71,72,109].

In the field of charging technologies, wired charging systems still dominate, offering the highest efficiency and broad compatibility with existing energy infrastructure [48,85]. However, wireless charging systems are gaining interest, particularly for urban fleets where convenience, automation, and minimal physical interaction become key advantages [16,83,85]. A particularly promising development direction is Vehicle-to-Grid (V2G) technology, which allows bidirectional energy flow and turns motorcycles into active participants in the power grid. Although this technology is still in the early stages of deployment, its potential for integration with renewable energy sources and local microgrids is significant [32,33,41].

Optimal configurations of electric motorcycle components, especially traction inverters, should be based on solutions that ensure high energy conversion efficiency, low switching losses, and resistance to electromagnetic interference [121]. In this context, modern inverters built using SiC transistors and multilevel architectures are particularly effective [27,53,54,55,56,57]. Contemporary design approaches also emphasize system modularity and the integration of ML-based diagnostic systems, which enable predictive fault detection, autonomous component switching, and drive configuration adjustment to operating conditions [58,59,60]. Furthermore, with the advancement of AI technologies and models, research is being conducted on the potential use of acoustic waves for extinguishing flames from various fuels and substances, which in the future may find application in both individual and public transport. Research in this area is the subject of numerous scientific efforts worldwide, e.g., [122,123,124,125,126,127,128,129,130,131].

In conclusion, the literature indicates that the future of electric motorcycles will depend on the synergistic development of high-performance PMSMs, advanced and safe energy storage technologies, flexible charging solutions, and ML-supported control systems. Thoughtful integration of these components into a coherent, scalable technical architecture will allow for the creation of more reliable, economical, and sustainable mobility solutions.

4. Results and Analysis

This chapter presents a detailed statistical analysis of 101 scientific publications in the field of electromobility, published across two five-year periods: 2015–2019 and 2020–2024. The analysis covered four main dimensions: document types (conference papers and journal articles), technological domains (electric drivetrains, energy storage, battery charging, electrical components), engineering and optimization topics (energy efficiency, sustainability, design and analysis), and applied methodological approaches (experiment, literature review, case study, conceptual approach, and survey-based research).

Additionally, a geographical comparison was conducted based on the authors’ country affiliations, and chi-square tests were applied to evaluate the statistical significance of observed differences. The results are presented in the form of tables, charts, and heatmaps, enabling the identification of dominant research trends, development directions, and areas requiring further investigation.

Based on the presented chart (Figure 5), it was shown that the number of conference papers increased from 21 in the period 2015–2019 to 48 in 2020–2024 (an increase of approximately 129%), while journal articles were recorded in quantities of 14 and 18, respectively (an increase of approximately 29%). The chart suggests that research findings were primarily disseminated through scientific conferences, highlighting the importance of rapid knowledge exchange. At the same time, the growing number of peer-reviewed journal articles indicates an increasing emphasis on formal validation of results.

The rise in the share of conference publications, alongside a moderate increase in journal articles, may be interpreted as a sign of intensified research dialog and a thematic shift from exploratory phases toward more established knowledge.

Figure 6 presents a comparison of the number of publications in four key areas of e-mobility technologies—electric drivetrains, energy storage, battery charging, and electrical components—across two consecutive five-year periods. In the area of electric drivetrains, the number of publications increased from 26 to 31 (an increase of approximately 19%). In the energy storage segment, the number of studies rose from 13 to 33 (approximately 154% growth). The most significant increase was observed in the field of battery charging, where the number of publications grew from 7 to 27 (an increase of approximately 286%). In contrast, the area of electrical components showed only a minimal change, with publications increasing from 9 to 10 (about 11% growth).

This publication dynamics structure is interpreted as a shift in research focus toward storage efficiency and fast charging, while interest in supporting components appears to have stabilized.

The marked growth in publications on battery charging between 2015 and 2019 and 2020–2024 reflects not only the expansion of charging infrastructure but also the rise in ML-assisted strategies for optimizing charging schedules, reducing costs under dynamic tariffs, and enabling V2G functionality [44,45,46,47,48,49]. In contrast, the modest increase in electrical component research suggests that while hardware improvements (e.g., SiC inverters) remain important, the integration of predictive diagnostics and adaptive control algorithms is an emerging but less explored area [53,54,55,56,57,58,59,60].

Figure 7 presents a comparison of the number of publications in three thematic categories related to engineering and optimization—energy efficiency, sustainability, and design and analysis—across the periods 2015–2019 and 2020–2024. In the first category, the number of studies increased from 15 to 41, representing a growth of approximately 173%. In the case of sustainability, no increase was recorded, with the number remaining constant at 20 (0% growth). In contrast, the design and analysis category saw an increase from 13 to 20 publications, corresponding to a growth of approximately 54%.

This trend indicates an intensification of research in energy efficiency and a strengthening of the role of design, while interest in sustainability-related issues appears to have stabilized.

The dominance of energy efficiency as a research theme aligns with the deployment of AI/ML models aimed at reducing energy losses in motors, optimizing energy flow in HESS configurations, and improving the accuracy of SoC/SoH estimation [17,32,33,34,40,46,70,72]. The stable number of sustainability-focused studies indicates that environmental lifecycle assessments are not growing at the same pace as performance-focused research, despite the potential of ML to optimize design for recyclability and eco-efficiency [58,59].

Figure 8 compares the number of publications across five research methodology categories—experimental studies, literature analysis, case studies, conceptual approaches, and survey-based research—across two consecutive five-year periods. For experimental methods, the number of publications increased from 18 to 38 (a growth of approximately 111%), while literature analysis rose from 9 to 25 (around 178%). In the case study category, the number of publications increased from 8 to 12 (50%), whereas conceptual approaches grew from 26 to 45 (about 73%). Only the survey category experienced a decline, dropping from 2 to 1 publication (a 50% decrease).

This pattern indicates a clear shift in emphasis toward experimental research and theoretical modeling, accompanied by a reduced reliance on survey-based methods.

The sharp increase in literature analysis publications corresponds to the systematic mapping of AI/ML applications across electric motorcycle subsystems, allowing researchers to identify high-impact algorithmic trends. The growth in experimental studies demonstrates a shift toward real-world validation of ML-based control and diagnostic systems, though survey-based research remains rare, limiting user feedback integration into AI-enabled designs.

Table 3. Publications by year in all categories.

Name	2015–2019	2020–2024	All Years	Share [%]
Total	35	66	101	100.0
Document Type
Conference Paper	21	48	69	68.32
Journal Article	14	18	32	31.68
E-mobility Technologies
Electric Motors	26	31	57	56.44
Energy Storage	13	33	46	45.54
Battery Charging	7	27	34	33.66
Electrical Components	9	10	19	18.81
Engineering and Optimization
Energy Efficiency	15	41	56	55.45
Sustainability	20	20	40	39.6
Design and Analysis	13	20	33	32.67
Research Methodology
Experiment	18	38	56	55.45
Literature Analysis	9	25	34	33.66
Case Study	8	12	20	19.8
Conceptual	26	47	73	72.28
Survey	2	1	3	2.97

presents a summary of 101 publications in the field of electromobility, divided into two time periods (2015–2019 and 2020–2024). The analysis covers four main dimensions: document type (Conference Paper vs. Journal Article), technological domains (Electric Motors, Energy Storage, Battery Charging, Electrical Components), optimization topics (Energy Efficiency, Sustainability, Design and Analysis), and research methodology (Experiment, Literature Analysis, Case Study, Conceptual, Survey).

In addition, statistical significance of differences between categories was assessed using the chi-square test within the groups “Technologies and Infrastructure” and “Research Methodology.”

Furthermore, chi-square statistical tests were applied across four groups: (1) Document Type, (2) E-mobility Technologies, (3) Engineering and Optimization, (4) Research Methodology.

Based on Table 3, a detailed, multidimensional analysis of the distribution of 101 scientific publications in the field of electromobility was carried out, covering two time periods: 2015–2019 (35 publications) and 2020–2024 (66 publications). The analysis included four main dimensions: document type, technological domain, optimization topics, and adopted research methodology. The obtained results were subjected to statistical verification using the chi-square test.

In the Document Type category, two groups were distinguished: Conference Paper and Journal Article. Conference papers accounted for 69 publications (68.32%), with 21 published before 2019 and 48 between 2020 and 2024. Journal articles were recorded in 32 cases (31.68%), including 14 in the first period and 18 in the second. It was observed that the share of journal articles increased by over 28% between the two periods, which may reflect the growing importance of peer review and the increasing stability of research findings.

The E-mobility Technologies analysis was divided into four subcategories: Electric Motors, Energy Storage, Battery Charging, and Electrical Components. Electric Motors were addressed in 57 publications (26 in 2015–2019; 31 in 2020–2024; 56.44%), indicating sustained interest in electric drive design and performance. Energy Storage appeared in 46 studies (13 + 33; 45.54%), with a significant increase in the second period, likely due to advancements in battery technologies. Battery Charging was the focus of 34 publications (7 + 27; 33.66%), reflecting growing challenges in charging infrastructure. Electrical Components were included in 19 studies (9 + 10; 18.81%), suggesting that hardware elements are less frequently explored, although they remain critical to system reliability.

In the Engineering and Optimization area, three key topics were identified: Energy Efficiency was covered in 56 publications (15 + 41; 55.45%), emphasizing the priority of optimizing energy consumption in EV systems. Sustainability appeared in 40 papers (20 + 20; 39.60%), indicating consistent interest in environmental aspects and component life cycles. Design and Analysis was the subject of 33 publications (13 + 20; 32.67%), highlighting the complexity of design and simulation processes necessary for developing safe and efficient solutions.

The Research Methodology distribution included five research approaches, with a dominance of experimental and conceptual studies: Experimental methods were used in 56 publications (18 + 38; 55.45%), reflecting high engagement in prototype testing and lab validation. Conceptual approaches were found in 73 publications (26 + 47; 72.28%), emphasizing the development of analytical and simulation-based models. Literature Analysis appeared in 34 papers (9 + 25; 33.66%), indicating strong foundations in review-based research. Case Studies were applied in 20 publications (8 + 12; 19.80%), while Survey-based studies were noted in only 3 cases (2 + 1; 2.97%), suggesting limited use of fieldwork and user feedback.

Chi-square tests were used to verify the significance of differences in four groups: for Document Type, the result was χ² = 1.17 (df = 1; p = 0.28); for E-mobility Technologies, χ² = 8.09 (df = 3; p = 0.04), indicating statistically significant differences between subcategories. In Engineering and Optimization, χ² = 5.47 (df = 2; p = 0.06); in Research Methodology, χ² = 2.78 (df = 4; p = 0.60), showing no statistically significant differences.

Key conclusions from the analysis include:

• Conference Papers dominate the electromobility research space, reflecting the role of rapid dissemination of preliminary results at conferences;
• Within E-mobility Technologies, Electric Motors and Energy Storage are the most frequently represented topics, while Electrical Components receive comparatively less attention;
• In Engineering and Optimization, Energy Efficiency holds central importance, reflecting the dominance of research focused on energy performance;
• In Research Methodology, experimental and conceptual studies account for more than two-thirds of all approaches, emphasizing the parallel development of practical testing and theoretical modeling.

Based on

Table 4. Publications by E-mobility Technologies in other categories.

Name	Electric Motors	Energy Storage	Battery Charging	Electrical Components	Total
Total	57	46	34	19	101
Engineering and Optimization
Energy Efficiency	27	27	20	10	56
Sustainability	21	15	14	5	40
Design and Analysis	25	13	6	11	33
Research Methodology
Experiment	31	28	19	13	56
Literature Analysis	21	14	10	8	34
Case Study	9	9	6	2	20
Conceptual	38	31	27	16	73
Survey	3	0	1	0	3

, “Publications by E-mobility Technologies in other categories,” a detailed analysis of the 101 scientific publications in the electromobility domain was conducted, with classification according to four technology categories: Electric Motors (57), Energy Storage (46), Battery Charging (34), and Electrical Components (19). Publications were assigned to two main thematic groups: Engineering and Optimization and Research Methodology.

Additionally, statistical analysis using the chi-square test was performed in two groups: (1) Engineering and Optimization, (2) Research Methodology.

In the Engineering and Optimization area, three subcategories were analyzed. For Energy Efficiency, 56 publications were identified, including 27 related to Electric Motors, 27 to Energy Storage, 20 to Battery Charging, and 10 to Electrical Components. The topic of Sustainability was addressed in 40 publications (21, 15, 14, and 5, respectively, for each of the listed technologies). Within Design and Analysis, 33 articles were identified, of which 25 focused on Electric Motors, 13 on Energy Storage, 6 on Battery Charging, and 11 on Electrical Components. This distribution indicates strong interest in optimizing drivetrains and storage systems, with moderate attention to design-related aspects and relatively lower engagement in research on electrical components.

In the Research Methodology group, five research approaches were compared. A total of 56 studies were classified under Experiment, of which 31 focused on Electric Motors, 28 on Energy Storage, 19 on Battery Charging, and 13 on Electrical Components. Literature Analysis included 34 publications (21, 14, 10, and 8 for the respective technologies).

Case Study was used in 20 studies, particularly involving Electric Motors and Battery Charging (9 and 6 cases, respectively), while Energy Storage appeared in 3 and Electrical Components in 2. Conceptual approaches were applied in 73 articles, with notable inclusion of Electric Motors (38), Energy Storage (31), Battery Charging (27), and Electrical Components (16). Survey appeared infrequently—in only 3 publications, primarily in the context of Electric Motors and Battery Charging. A clear dominance of Conceptual and Experimental approaches was observed, while Survey and Case Study methods were used less frequently. For the Engineering and Optimization group, the chi-square test yielded χ² = 8.69 (df = 6; p = 0.19), indicating no statistically significant differences in the distribution of publications across technological subcategories. Similarly, in the Research Methodology group, the result was χ² = 5.78 (df = 12; p = 0.93), confirming no significant differences in the choice of research methods.

From the analysis, the following conclusions can be drawn:

• Electric Motors technology was the most frequently addressed in both optimization and methodological studies, which may indicate the central role of drivetrains in the development of electromobility.
• Energy Storage ranked just behind drivetrains, underscoring the importance of storage systems in EV design.
• Battery Charging and Electrical Components were represented to a lesser extent, suggesting the need for further research in charging infrastructure and electronic componentry.
• The absence of statistically significant differences suggests a relatively even distribution of research interest across technologies in the context of both optimization and methodology.

Figure 9 presents a comparison of the number of publications across two five-year periods, categorized by the authors’ country of affiliation. For Indonesia, the number of publications increased from 2 to 18, representing the most significant change in the dataset. In the case of Japan, the number of studies remained constant at 5, while in the United Kingdom, it rose from 3 to 6. The United States saw a modest increase from 3 to 4 publications, while Canada and Thailand maintained stable levels of 3 and 3, respectively. Italy’s publication count increased from 2 to 4, and India recorded a rise from 0 to 5. A decrease from 3 to 2 publications was registered for Malaysia, while both South Korea and Spain experienced an increase from 2 to 3 publications. For the “Other” category, the number of studies rose from 9 to 15.

This dynamic indicates a shift in research activity toward regions with a growing presence in the field of electromobility literature.

Based on

Table 5. Publications by year in Countries.

Country	2015–2019	2020–2024	All Years	Share [%]
All countries	35	66	101	100.0
Indonesia	2	18	20	19.8
Japan	5	5	10	9.9
United Kingdom	3	6	9	8.91
United States	3	4	7	6.93
Canada	3	3	6	5.94
Italy	2	4	6	5.94
Thailand	3	3	6	5.94
India	0	5	5	4.95
Malaysia	3	2	5	4.95
South Korea	2	3	5	4.95
Spain	2	3	5	4.95
Other	9	15	24	23.76

, the distribution of 101 publications in the field of electromobility is presented according to the authors’ country of affiliation across the two periods (2015–2019 and 2020–2024), showing each country’s percentage share and the result of a chi-square test verifying the significance of observed changes.

To assess whether the changes in the number of publications across individual countries between the periods 2015–2019 and 2020–2024 were statistically significant, a chi-square test was applied.

• Chi-square value (χ²): 12.22
• Degrees of freedom: 11
• p-value: 0.35

Since the p-value exceeded the significance threshold of 0.05, no statistically significant association was found between the year of publication and the authors’ country of affiliation.

The analysis included 101 publications, with 35 published in 2015–2019 and 66 in 2020–2024. The largest share of studies was attributed to Indonesia (20; 19.8%), where the number of publications increased from 2 to 18 across the two timeframes. Japan ranked second with an even distribution (5 + 5; 9.9%), while the United Kingdom followed in third place (3 + 6; 8.91%). The United States had 7 publications (3 + 4; 6.93%), and Canada published 6 (3 + 3; 5.94%), reflecting the relative stability of their contributions. A similar number of studies (6; 5.94%) was observed in Italy and Thailand, while India, Malaysia, South Korea, and Spain each accounted for 5 publications (4.95%). The “Other” category, comprising 24 papers (23.76%), indicates significant involvement from smaller research centers.

The analysis leads to the following conclusions:

• Indonesia emerged as the leading research hub during the analyzed period, accounting for nearly 20% of all publications.
• Japan and the UK maintained steady levels of research activity, possibly reflecting stable research funding.
• The increase in India’s output from 0 to 5 publications suggests a dynamic growth in scientific engagement in South Asia.
• Countries with a moderate contribution, such as Canada, Italy, and Thailand, maintained a consistent level of output, possibly due to established research strategies.
• The high proportion of papers classified under “Other” reflects growing geographical diversity and increasing activity of smaller research institutions worldwide—possibly the result of intentional research policy initiatives.

The presented heatmap (Figure 10) illustrates the distribution of publications by research methodology (experiment, literature review, case study, conceptual approach, survey) in the context of four e-mobility technologies (electric motors, energy storage, battery charging, electrical components).

The most frequently employed research methodology was experimental studies, highlighting the importance of practical validation of technological solutions. This group included:

• 31 publications related to electric motors,
• 28 analyses of energy storage,
• 19 studies focused on battery charging,
• 13 works addressing electrical components.

Literature analysis was identified as the second most commonly used method, with 21 publications on electric motors, 14 on energy storage, 10 on battery charging, and 8 on electrical components. Case studies were relatively rare—conducted in 9 studies on motors and storage, 6 on charging, and only 2 on components—indicating limited documentation of real-world implementations. Conceptual approaches proved essential for theoretical development, as evidenced by 38 publications on electric motors, 31 on energy storage, 27 on battery charging, and 16 on electrical components. Surveys were used sporadically—3 times in the context of motors, once in charging, and none for storage or components.

This methodological structure is considered to reflect the advanced stage of e-mobility technology development, where practical testing and conceptual framework development go hand in hand. The limited number of case studies and surveys suggests a gap, particularly in documenting implementations and gathering user data. Literature analysis, meanwhile, is seen as the foundation for building a robust theoretical base, enabling the generation of new hypotheses. The presented distribution may serve to guide future research directions by identifying both methodological and technological gaps.

The presented heatmap (Figure 11) illustrates the relationship between three categories of engineering and optimization topics and four e-mobility technologies. Each cell contains the number of publications in which a given topic (energy efficiency, sustainability, design and analysis) is linked to a specific technology (electric motors, energy storage, battery charging, electrical components). This enables the identification of the most frequently explored thematic combinations in the reviewed literature.

Energy efficiency was identified as the most frequently represented category within engineering and optimization, highlighting its central role in e-mobility research. This group included:

• 27 publications related to electric motors,
• 27 studies on energy storage,
• 20 papers on battery charging,
• 10 publications concerning electrical components.

Such high values were interpreted as a concentration of scientific efforts on optimizing energy consumption in the core elements of EV systems, with gradually decreasing attention given to auxiliary components.

Sustainability was the second most prominent category, with:

• 21 publications focused on electric motors,
• 15 on energy storage,
• 14 on battery charging,
• 5 related to electrical components.

The moderate level of attention given to charging and components was seen as a signal of the need for deeper investigation into the environmental impact and life cycle of e-mobility technologies.

In the area of design and analysis, the following numbers were observed:

• 25 publications on electric motors,
• 13 on energy storage,
• 6 on battery charging,
• 11 on electrical components.

This varied distribution reflects the growing importance of interdisciplinary simulation and modeling methods in the integration of different technologies.

In summary, energy efficiency clearly dominated the reviewed literature, while interest in sustainability and in design and analysis was more evenly balanced across studies on electric motors. The least attention was paid to electrical components.

The analysis of 101 scientific publications in the field of electromobility revealed a clear increase in research activity in the period 2020–2024 compared to 2015–2019. Conference papers remained the dominant form of publication, underscoring the importance of rapid dissemination in a rapidly evolving field. At the same time, the growing share of peer-reviewed journal articles reflects the scientific community’s pursuit of formal validation and stabilization of results.

Among the technological areas, electric drivetrains and energy storage attracted the greatest interest. The most dynamic growth in the number of publications was observed in the field of battery charging, likely due to increasing demands on charging infrastructure. Electrical components, while essential for EV system reliability, were relatively less represented in the literature.

Engineering-related topics focused primarily on energy efficiency, appearing in over half of the analyzed studies. Although sustainability and design/analysis were also present, they appeared to a lesser extent. This distribution suggests a practical orientation in research, with energy savings being treated as a top priority in e-mobility systems.

From a methodological perspective, experimental and conceptual approaches were dominant, indicating simultaneous development of both practical solutions and theoretical models. Literature reviews and case studies played a supplementary role, while survey-based research was marginal. This methodological structure suggests that research in this domain is still rarely conducted with direct user participation or in the context of real-world implementations.

Geographically, the largest increase in the number of publications was observed in Indonesia, which emerged as the leading contributor to electromobility research. Japan, the United Kingdom, and the United States maintained consistent levels of activity. At the same time, the number of publications from previously less-represented countries increased, suggesting growing internationalization of research and rising interest in e-mobility across developing regions.

In conclusion, this chapter reveals the dynamic scientific development in the field of electromobility, with a clear shift in research focus toward energy storage, battery charging, and energy efficiency. The dominance of experimental and conceptual methods, along with the limited use of applied approaches, points to the potential for further exploration in areas related to implementation and end-user perspectives.

5. Discussion and Future Work

Table 6. Summary of ML-driven trends, benefits, challenges, and future directions across technology domains.

Domain	Key ML Algorithms	Reported Benefits	Main Challenges	Future Directions
Electric Motors	LSTM, ANFIS, RL, SVM	+3–5% efficiency, torque ripple reduction, improved adaptability	Computational load, large datasets required	Lightweight hybrid control models for embedded systems
Energy Storage	CNN, Ensemble, GRU, Kalman filters	SoC/SoH error ≤ 2%, extended battery lifespan	Sensor noise, temperature sensitivity	Federated BMS learning, thermal-aware predictive models
Charging and V2G	RL, LSTM, Boosting, SVM	Reduced charging cost/time, improved grid stability	Infrastructure readiness, interoperability issues	Hybrid scheduling combining V2G, RES, and HESS
Electronic Components	Autoencoders, CNN, Decision Trees	Early fault detection, reduced downtime	Generalization across platforms, limited datasets	XAI-based diagnostics, adaptive fault-tolerant control

summarizes the most relevant machine learning (ML) trends identified across the four main technological domains of electric motorcycles. For each domain, key ML algorithms are listed alongside reported performance benefits, major technical challenges, and suggested future research directions. The synthesis highlights that, while ML has achieved measurable gains—such as improved torque control in motors, higher accuracy in battery state estimation, and cost-optimized charging—challenges remain in terms of computational efficiency, model generalization, and infrastructure readiness. Addressing these gaps, particularly through lightweight models, federated learning, and Explainable AI (XAI), will be essential for large-scale deployment in commercial electric motorcycle systems.

The comparative overview in

Table 7. Comparative performance of major ML algorithms in the four key technological domains.

Domain	ML Algorithms	Application	Measured Effect	Representative Sources
Electric Motors	LSTM, ANFIS, RL, SVM	Torque control, ripple reduction, load adaptation	+3–5% efficiency, 8–12% torque ripple reduction, 10–15% faster dynamic response	[14,17,20,21,22,23,24]
Energy Storage	CNN, Ensemble, GRU, Kalman filters	SoC/SoH estimation, degradation forecasting	≤2% SoC error, 5–10% longer lifespan, >90% fault prediction accuracy	[32,33,34,46,53,58,59]
Charging and V2G	RL, LSTM, Boosting, SVM	Cost/time optimization, renewable integration	10–15% faster charging, 5–12% cost reduction, 8–10% peak load reduction	[16,44,46,47,48,49]
Electronic Components	Autoencoders, CNN, Decision Trees	Inverter diagnostics, fault prediction	>95% fault detection accuracy, 10–20% downtime reduction	[53,54,55,56,57,58,59,60]

reveals several key cross-domain patterns:

• AI/ML penetration is highest in control and prediction tasks. PMSMs and energy storage show the most mature applications, where LSTM, ANFIS, and CNN architectures deliver measurable performance gains.
• Data quality and computational constraints remain universal bottlenecks, limiting real-time deployment in embedded systems.
• Charging and V2G applications are emerging as high-growth areas, but infrastructure readiness and interoperability remain barriers.
• Electronic components lag in research intensity, although ML-based diagnostics have demonstrated clear benefits in fault detection and maintenance reduction.
• Future directions converge toward integrated, lightweight, and explainable models, suggesting a shift from isolated component optimization toward fully connected, AI-driven motorcycle architectures.

As summarized in Table 6, the systematic literature review conducted for the period 2015–2024 revealed that electric motorcycles are becoming an increasingly important component of electromobility, with technological development focused around four key areas: electric motors, energy storage, charging systems, and electronic components. Across these domains, the integration of artificial intelligence (AI) and machine learning (ML) has emerged as a unifying factor, enhancing control precision, predictive maintenance capabilities, and overall system efficiency.

5.1. Key Findings and Discussion

In the context of electric motor efficiency, Permanent Magnet Synchronous Motors (PMSMs) remain the most attractive option in terms of energy performance and controllability, particularly when combined with ML-enhanced control strategies such as LSTM-assisted Field-Oriented Control and adaptive neuro-fuzzy inference systems. These approaches have been shown to reduce torque ripple, improve dynamic response, and adapt to variable load conditions better than conventional methods. Brushless DC motors (BLDC) and Switched Reluctance Motors (SRM) remain relevant for cost-sensitive applications, where ML-based optimization can mitigate their inherent disadvantages, such as acoustic noise and complex torque profiles.

For energy storage technologies, lithium-ion cells continue to dominate due to their favorable capacity-to-weight ratio and thermal performance. However, ML-supported Battery Management Systems (BMS) now play a key role in maximizing their lifespan and reliability, providing accurate state-of-charge (SoC) and state-of-health (SoH) estimation, as well as adaptive control to mitigate degradation. Hybrid Energy Storage Systems (HESS), integrating batteries with supercapacitors, further benefit from reinforcement learning and predictive algorithms to optimize energy flow during acceleration and regenerative braking.

Regarding charging technologies, wired systems currently offer the highest efficiency and compatibility, but wireless and Vehicle-to-Grid (V2G) solutions are gaining traction. In particular, ML-based charging optimization—using reinforcement learning, LSTM forecasting, and boosting algorithms—enables intelligent scheduling under dynamic tariffs, integration with renewable energy sources, and grid load balancing. Although V2G for motorcycles remains at an early stage, predictive and adaptive algorithms hold significant potential for future deployment.

For electronic components, modern traction inverters built with SiC transistors and multilevel architectures are delivering higher conversion efficiencies and lower switching losses. ML-driven diagnostics embedded in these components allow for early fault detection, autonomous component switching, and adaptive performance tuning, directly improving system reliability and reducing downtime.

5.2. Future Research Directions

Future research should focus on the following strategic directions:

• Integrated AI/ML Architectures—Developing unified control frameworks that combine motor drive optimization, energy storage management, charging control, and fault diagnostics within a single intelligent system.
• Lightweight and Energy-Efficient Models—Designing ML algorithms optimized for embedded controllers with strict computational and power constraints, enabling real-time operation in edge environments.
• Explainable AI (XAI)—Increasing the interpretability of ML models to improve user trust, support certification processes, and ensure safe deployment in safety-critical functions.
• Federated Learning and Privacy-Preserving Systems—Enabling collaborative training of ML models across fleets without transmitting raw operational data, thereby enhancing security and scalability.
• Advanced Hybrid Energy Management—Integrating fast charging, V2G capabilities, and HESS within a single ML-optimized framework to improve flexibility, sustainability, and cost-effectiveness.
• Eco-Design Integration—Combining sustainable materials and lightweight construction with intelligent control systems to reduce lifecycle environmental impact. Standardized Performance Benchmarks—Establishing common metrics for evaluating ML-based solutions in electric motorcycles to facilitate cross-study comparisons and industrial adoption.

By pursuing these directions, future electric motorcycles can evolve into fully intelligent platforms capable of continuous self-optimization, predictive adaptation to environmental and operational conditions, and seamless integration with the wider energy ecosystem. This convergence of hardware innovation and AI/ML-driven intelligence represents a decisive step toward achieving the twin goals of high performance and environmental sustainability in the electromobility sector.

6. Conclusions

This review has provided a comprehensive analysis of technological advancements in electric motorcycles between 2015 and 2024, focusing on four main domains: electric motors, energy storage, charging technologies, and electronic components. As highlighted in Table 6 and further discussed in the preceding section, the integration of artificial intelligence (AI) and machine learning (ML) into these domains is no longer a peripheral trend but a central driver of innovation.

The findings confirm that PMSMs deliver the highest efficiency and controllability among electric motor types, especially when combined with ML-enhanced control strategies. Lithium-ion batteries remain the dominant energy storage technology, with ML-supported Battery Management Systems significantly improving state estimation accuracy and extending service life. Hybrid Energy Storage Systems (HESSs) demonstrate strong potential for enhancing performance, particularly when paired with predictive ML-based energy flow control. Charging technologies are evolving beyond traditional wired solutions toward intelligent, ML-driven wireless and Vehicle-to-Grid (V2G) systems, enabling adaptive integration with renewable energy sources. Modern electronic components—such as SiC-based inverters—combined with embedded ML diagnostics are improving fault detection, operational safety, and system adaptability.

Across all domains, ML enables higher efficiency, predictive capabilities, and adaptive control; however, most implementations remain focused on individual subsystems rather than integrated architectures. The absence of standardized performance benchmarks limits comparability and hinders large-scale adoption.

The future trajectory of electric motorcycle development will depend on advancing integrated AI/ML architectures, deploying lightweight and energy-efficient models for embedded systems, implementing Explainable AI for transparency and trust, and combining eco-design principles with intelligent energy management. These advancements will be critical to achieving the twin objectives of high performance and environmental sustainability.

In conclusion, the convergence of hardware innovation and AI/ML-driven intelligence represents a transformative shift in electric motorcycle design. By uniting advances in motor technology, energy storage, charging systems, and electronic components under an integrated, intelligent framework, the industry can accelerate the transition toward smarter, more efficient, and more sustainable mobility solutions.