Robotics

34 pages, 2660 KiB

Open AccessArticle

Cascade-Based Distributed Estimator Tracking Control for Swarm of Multiple Nonholonomic Wheeled Mobile Robots via Leader–Follower Approach

by Dinesh Elayaperumal, Sachin Sakthi Kuppusami Sakthivel, Sathishkumar Moorthy, Sathiyamoorthi Arthanari, Young Hoon Joo and Jae Hoon Jeong

Robotics 2025, 14(7), 88; https://doi.org/10.3390/robotics14070088 (registering DOI) - 26 Jun 2025

Abstract

This study aims to explore the tracking control challenge in a swarm of multiple nonholonomic wheeled mobile robots (NWMRs) by utilizing a distributed leader–follower strategy grounded in the cascade system theory. Firstly, the kinematic control law is developed for the leader by constructing [...] Read more.

This study aims to explore the tracking control challenge in a swarm of multiple nonholonomic wheeled mobile robots (NWMRs) by utilizing a distributed leader–follower strategy grounded in the cascade system theory. Firstly, the kinematic control law is developed for the leader by constructing a sliding surface based on the error tracking model with a virtual reference trajectory. Secondly, a communication topology with the desired formation pattern is modeled for the multiple robots by using the graph theory. Further, in the leader–follower NWMR system, each follower lacks direct access to the leader’s information. Therefore, a novel distributed-based controller by PD-based controller for the follower is developed, enabling each follower to obtain the leader’s information. Thirdly, for each case, we give a further analysis of the closed-loop system to guarantee uniform global asymptotic stability with the conditions based on the cascade system theory. Finally, the trajectory tracking performance of the proposed controllers for the NWMR system is illustrated through simulation results. The leader robot achieved a low RMSE of 1.6572 (Robot 1), indicating accurate trajectory tracking. Follower robots showed RMSEs of 2.6425 (Robot 2), 3.0132 (Robot 3), and 4.2132 (Robot 3), reflecting minor variations due to the distributed control strategy and local disturbances. Full article

(This article belongs to the Section Sensors and Control in Robotics)

► Show Figures

Figure 1

18 pages, 24410 KiB

Open AccessArticle

Design and Experimental Validation of a 3D-Printed Two-Finger Gripper with a V-Shaped Profile for Lightweight Waste Collection

by Mahboobe Habibi, Giuseppe Sutera, Dario Calogero Guastella and Giovanni Muscato

Robotics 2025, 14(7), 87; https://doi.org/10.3390/robotics14070087 (registering DOI) - 25 Jun 2025

Abstract

This study presents the design, fabrication, and experimental validation of a two-finger robotic gripper featuring a 135

^{\circ}

V-shaped fingertip profile tailored for lightweight waste collection in laboratory-scale environmental robotics. The gripper was developed with a strong emphasis on cost-effectiveness and manufacturability, utilizing [...] Read more.

This study presents the design, fabrication, and experimental validation of a two-finger robotic gripper featuring a 135

^{\circ}

V-shaped fingertip profile tailored for lightweight waste collection in laboratory-scale environmental robotics. The gripper was developed with a strong emphasis on cost-effectiveness and manufacturability, utilizing a desktop 3D printer and off-the-shelf servomotors. A four-bar linkage mechanism enables parallel jaw motion and ensures stable surface contact during grasping, achieving a maximum opening range of 71.5 mm to accommodate common cylindrical objects. To validate structural integrity, finite element analysis (FEA) was conducted under a 0.6 kg load, yielding a safety factor of 3.5 and a peak von Mises stress of 12.75 MPa—well below the material yield limit of PLA. Experimental testing demonstrated grasp success rates of up to 80 percent for typical waste items, including bottles, disposable cups, and plastic bags. While the gripper performs reliably with rigid and semi-rigid objects, further improvements are needed for handling highly deformable materials such as thin films or soft bags. The proposed design offers significant advantages in terms of rapid prototyping (a print time of approximately 10 h), modularity, and low manufacturing cost (with an estimated in-house material cost of USD 20 to 40). It provides a practical and accessible solution for small-scale robotic waste-collection tasks and serves as a foundation for future developments in affordable, application-specific grippers. Full article

(This article belongs to the Section Intelligent Robots and Mechatronics)

18 pages, 14780 KiB

Open AccessArticle

Boosting Deep Reinforcement Learning with Semantic Knowledge for Robotic Manipulators

by Lucía Güitta-López, Vincenzo Suriani, Jaime Boal, Álvaro J. López-López and Daniele Nardi

Robotics 2025, 14(7), 86; https://doi.org/10.3390/robotics14070086 - 24 Jun 2025

Abstract

Deep Reinforcement Learning (DRL) is a powerful framework for solving complex sequential decision-making problems, particularly in robotic control. However, its practical deployment is often hindered by the substantial amount of experience required for learning, which results in high computational and time costs. In [...] Read more.

Deep Reinforcement Learning (DRL) is a powerful framework for solving complex sequential decision-making problems, particularly in robotic control. However, its practical deployment is often hindered by the substantial amount of experience required for learning, which results in high computational and time costs. In this work, we propose a novel integration of DRL with semantic knowledge in the form of Knowledge Graph Embeddings (KGEs), aiming to enhance learning efficiency by providing contextual information to the agent. Our architecture combines KGEs with visual observations, enabling the agent to exploit environmental knowledge during training. Experimental validation with robotic manipulators in environments featuring both fixed and randomized target attributes demonstrates that our method achieves up to 60% reduction in learning time and improves task accuracy by approximately 15 percentage points, without increasing training time or computational complexity. These results highlight the potential of semantic knowledge to reduce sample complexity and improve the effectiveness of DRL in robotic applications. Full article

(This article belongs to the Special Issue Applications of Neural Networks in Robot Control)

► Show Figures

Figure 1

25 pages, 3468 KiB

Open AccessArticle

Distributed Monitoring of Moving Thermal Targets Using Unmanned Aerial Vehicles and Gaussian Mixture Models

by Yuanji Huang, Pavithra Sripathanallur Murali and Gustavo Vejarano

Robotics 2025, 14(7), 85; https://doi.org/10.3390/robotics14070085 - 22 Jun 2025

Abstract

This paper contributes a two-step approach to monitor clusters of thermal targets on the ground using unmanned aerial vehicles (UAVs) and Gaussian mixture models (GMMs) in a distributed manner. The approach is tailored to networks of UAVs that establish a flying ad hoc [...] Read more.

This paper contributes a two-step approach to monitor clusters of thermal targets on the ground using unmanned aerial vehicles (UAVs) and Gaussian mixture models (GMMs) in a distributed manner. The approach is tailored to networks of UAVs that establish a flying ad hoc network (FANET) and operate without central command. The first step is a monitoring algorithm that determines if the GMM corresponds to the current spatial distribution of clusters of thermal targets on the ground. UAVs make this determination using local data and a sequence of data exchanges with UAVs that are one-hop neighbors in the FANET. The second step is the calculation of a new GMM when the current GMM is found to be unfit, i.e., the GMM no longer corresponds to the new distribution of clusters on the ground due to the movement of thermal targets. A distributed expectation-maximization algorithm is developed for this purpose, and it operates on local data and data exchanged with one-hop neighbors only. Simulation results evaluate the performance of both algorithms in terms of the number of communication exchanges. This evaluation is completed for an increasing number of clusters of thermal targets and an increasing number of UAVs. The performance is compared with well-known solutions to the monitoring and GMM calculation problems, demonstrating convergence with a lower number of communication exchanges. Full article

(This article belongs to the Special Issue Multi-Robot Systems for Environmental Monitoring and Intervention)

► Show Figures

Figure 1

43 pages, 10203 KiB

Open AccessArticle

Neural Adaptive Nonlinear MIMO Control for Bipedal Walking Robot Locomotion in Hazardous and Complex Task Applications

by Belkacem Bekhiti, Jamshed Iqbal, Kamel Hariche and George F. Fragulis

Robotics 2025, 14(6), 84; https://doi.org/10.3390/robotics14060084 - 17 Jun 2025

Abstract

This paper introduces a robust neural adaptive MIMO control strategy to improve the stability and adaptability of bipedal locomotion amid uncertainties and external disturbances. The control combines nonlinear dynamic inversion, finite-time convergence, and radial basis function (RBF) neural networks for fast, accurate trajectory [...] Read more.

This paper introduces a robust neural adaptive MIMO control strategy to improve the stability and adaptability of bipedal locomotion amid uncertainties and external disturbances. The control combines nonlinear dynamic inversion, finite-time convergence, and radial basis function (RBF) neural networks for fast, accurate trajectory tracking. The main novelty of the presented control strategy lies in unifying instantaneous feedback, real-time learning, and dynamic adaptation within a multivariable feedback framework, delivering superior robustness, precision, and real-time performance under extreme conditions. The control scheme is implemented on a 5-DOF underactuated

R A B B I T

robot using a

d S P A C E D S 1103

platform with a sampling rate of

∆ t = 1.5 m s (667 H z)

. The experimental results show excellent performance with the following: The robot achieved stable cyclic gaits while keeping the tracking error within

e = \pm 0.04 r a d

under nominal conditions. Under severe uncertainties of trunk mass variations

{∆ m}_{t r u n k} = + 100 %

, limb inertia changes

{∆ I}_{l i m b} = \pm 30 %

, and actuator torque saturation at

τ = \pm 150 N m

, the robot maintains stable limit cycles with smooth control. The performance of the proposed controller is compared with classical nonlinear decoupling, non-adaptive finite-time, neural-fuzzy learning, and deep learning controls. The results demonstrate that the proposed method outperforms the four benchmark strategies, achieving the lowest errors and fastest convergence with the following:

I A E = 1.36

,

I T A E = 2.43

,

I S E = 0.68

,

t_{s s} = 1.24 s

, and

M_{p} = 2.21 %

. These results demonstrate evidence of high stability, rapid convergence, and robustness to disturbances and foot-slip. Full article

(This article belongs to the Section Humanoid and Human Robotics)

► Show Figures

Figure 1

23 pages, 4949 KiB

Open AccessArticle

Hybrid LDA-CNN Framework for Robust End-to-End Myoelectric Hand Gesture Recognition Under Dynamic Conditions

by Hongquan Le, Marc in het Panhuis, Geoffrey M. Spinks and Gursel Alici

Robotics 2025, 14(6), 83; https://doi.org/10.3390/robotics14060083 - 17 Jun 2025

Abstract

Gesture recognition based on conventional machine learning is the main control approach for advanced prosthetic hand systems. Its primary limitation is the need for feature extraction, which must meet real-time control requirements. On the other hand, deep learning models could potentially overfit when [...] Read more.

Gesture recognition based on conventional machine learning is the main control approach for advanced prosthetic hand systems. Its primary limitation is the need for feature extraction, which must meet real-time control requirements. On the other hand, deep learning models could potentially overfit when trained on small datasets. For these reasons, we propose a hybrid Linear Discriminant Analysis–convolutional neural network (LDA-CNN) framework to improve the gesture recognition performance of sEMG-based prosthetic hand control systems. Within this framework, 1D-CNN filters are trained to generate latent representation that closely approximates Fisher’s (LDA’s) discriminant subspace, constructed from handcrafted features. Under the train-one-test-all evaluation scheme, our proposed hybrid framework consistently outperformed the 1D-CNN trained with cross-entropy loss only, showing improvements from 4% to 11% across two public datasets featuring hand gestures recorded under various limb positions and arm muscle contraction levels. Furthermore, our framework exhibited advantages in terms of induced spectral regularization, which led to a state-of-the-art recognition error of 22.79% with the extended 23 feature set when tested on the multi-limb position dataset. The main novelty of our hybrid framework is that it decouples feature extraction in regard to the inference time, enabling the future incorporation of a more extensive set of features, while keeping the inference computation time minimal. Full article

(This article belongs to the Special Issue AI for Robotic Exoskeletons and Prostheses)

► Show Figures

Figure 1

20 pages, 8680 KiB

Open AccessArticle

Humanoid Motion Generation in Complex 3D Environments

by Diego Marussi, Michele Cipriano, Nicola Scianca, Leonardo Lanari and Giuseppe Oriolo

Robotics 2025, 14(6), 82; https://doi.org/10.3390/robotics14060082 - 16 Jun 2025

Abstract

We address the problem of humanoid locomotion in 3D environments consisting of planar regions with arbitrary inclination and elevation, such as staircases, ramps, and multi-floor layouts. The proposed framework combines an offline randomized footstep planner with an online control pipeline that includes a [...] Read more.

We address the problem of humanoid locomotion in 3D environments consisting of planar regions with arbitrary inclination and elevation, such as staircases, ramps, and multi-floor layouts. The proposed framework combines an offline randomized footstep planner with an online control pipeline that includes a model predictive controller for gait generation and a whole-body controller for computing robot torque commands. The planner efficiently explores the environment and returns the highest-quality plan it can find within a user-specified time budget, while the control layer ensures dynamic balance and adequate ground friction. The complete framework was evaluated via dynamic simulation in MuJoCo, placing the JVRC1 humanoid in four scenarios of varying complexity. Full article

(This article belongs to the Section Humanoid and Human Robotics)

► Show Figures

Figure 1

27 pages, 3094 KiB

Open AccessReview

Innovations in Upper Limb Rehabilitation Robots: A Review of Mechanisms, Optimization, and Clinical Applications

by Yang Wang, Xu Han, Baiye Xin and Ping Zhao

Robotics 2025, 14(6), 81; https://doi.org/10.3390/robotics14060081 - 11 Jun 2025

Abstract

With the continuous increase in the global aging population, stroke has become one of the major diseases affecting the health of the elderly, and the upper limb motor dysfunction it causes often requires long-term rehabilitation. To improve rehabilitation outcomes for hemiplegic patients and [...] Read more.

With the continuous increase in the global aging population, stroke has become one of the major diseases affecting the health of the elderly, and the upper limb motor dysfunction it causes often requires long-term rehabilitation. To improve rehabilitation outcomes for hemiplegic patients and alleviate the shortage of rehabilitation physicians, upper limb rehabilitation robots have shown great potential in enhancing motor function and improving stroke patients’ rehabilitation outcomes in clinical research. This paper first classifies rehabilitation robots based on their driving mechanisms and interaction modes, describing the application of their structural features in various scenarios. It then analyzes the optimization methods used in the trajectory planning process of rehabilitation robots at different stages. Finally, based on existing shortcomings, the paper summarizes the future development directions of upper limb rehabilitation robots, providing prospects for the development of upper limb rehabilitation robots in the areas of artificial intelligence and compliant control, multi-sensory feedback and interactive training, ergonomics and new driving technologies, modular and customizable designs, and multi-modal brain stimulation techniques. Full article

(This article belongs to the Section Medical Robotics and Service Robotics)

► Show Figures

Figure 1

18 pages, 6883 KiB

Open AccessArticle

Autonomous, Collaborative, and Confined Infrastructure Assessment with Purpose-Built Mega-Joey Robots

by Hitesh Bhardwaj, Nabil Shaukat, Andrew Barber, Andy Blight, George Jackson-Mills, Andrew Pickering, Manman Yang, Muhammad Azam Mohd Sharif, Linyan Han, Songyan Xin and Robert Richardson

Robotics 2025, 14(6), 80; https://doi.org/10.3390/robotics14060080 - 10 Jun 2025

Abstract

The inspection of sewer pipes in the UK is costly, and if not inspected regularly, they are costly and disruptive to repair. This paper presents the Mega-Joey, a novel miniature, tether-less robot platform that is capable of autonomously navigating and assessing confined spaces, [...] Read more.

The inspection of sewer pipes in the UK is costly, and if not inspected regularly, they are costly and disruptive to repair. This paper presents the Mega-Joey, a novel miniature, tether-less robot platform that is capable of autonomously navigating and assessing confined spaces, such as small-diameter underground pipelines. This paper also discusses a novel decentralized event-based-broadcasting autonomous exploration algorithm designed for exploring such pipe networks collaboratively. The designed robot is able to operate in pipes with an inclination of up to 20 degrees in dry and up to 10 degrees in wet conditions. A team of Mega-Joeys was used to explore a test network using the proposed algorithm. The experimental results show that the team of robots was able to explore a 3850 mm long test network within a faster period (36% faster) and in a more energy-efficient manner (approximately 54% more efficient) than a single robot could achieve. Full article

(This article belongs to the Section Intelligent Robots and Mechatronics)

► Show Figures

Figure 1

21 pages, 5263 KiB

Open AccessArticle

Design and Analysis of an Adaptable Wheeled-Legged Robot for Vertical Locomotion

by Ernesto Christian Orozco-Magdaleno, Eduardo Castillo-Castañeda, Omar Rodríguez-Abreo and Giuseppe Carbone

Robotics 2025, 14(6), 79; https://doi.org/10.3390/robotics14060079 - 10 Jun 2025

Abstract

Most of the developed and studied service robots for vertical locomotion, as visual inspection, are made up by a rigid body with legs, wheels, or both. Thus, the robot can only displace over regular and/or flat surfaces since it is not able to [...] Read more.

Most of the developed and studied service robots for vertical locomotion, as visual inspection, are made up by a rigid body with legs, wheels, or both. Thus, the robot can only displace over regular and/or flat surfaces since it is not able to adapt to the irregularities and projections of the wall. Therefore, this paper presents the design and analysis of an adaptable robot for vertical locomotion service tasks, which has a body made up of four wheeled legs that can easily adapt to the different irregularities and projections of building facades. The robot uses an Electric Ducted Fan (EDF) as the vortex adhesion system. Each leg has a rubber cover, which allows a higher mechanical adaptability of the robot over different irregularities of the wall. Theoretical backgrounds and open issues are addressed by considering some challenging problems such as mechanical adaptability modeling as well as kinematic and static analysis. Laser sensors are mounted over the robot to measure the adaptability of the robot, between the legs and body, at each time of the experimental tests for vertical locomotion. Full article

(This article belongs to the Special Issue Legged Robots into the Real World, 2nd Edition)

► Show Figures

Figure 1

28 pages, 7611 KiB

Open AccessArticle

Design and Experimental Study of a Robotic System for Target Point Manipulation in Breast Procedures

by Bing Li, Hafiz Muhammad Muzzammil, Junwu Zhu and Lipeng Yuan

Robotics 2025, 14(6), 78; https://doi.org/10.3390/robotics14060078 - 2 Jun 2025

Abstract

To achieve obstacle-avoiding puncture in breast interventional surgery, a robotics system based on three-fingered breast target-point manipulation is proposed and designed. Firstly, based on the minimum number of control points required for three-dimensional breast deformation control and the bionic structure of the human [...] Read more.

To achieve obstacle-avoiding puncture in breast interventional surgery, a robotics system based on three-fingered breast target-point manipulation is proposed and designed. Firstly, based on the minimum number of control points required for three-dimensional breast deformation control and the bionic structure of the human hand, the structure and control scheme of the robotics system based on breast target-point manipulation are proposed. Additionally, the workspace of the robotics system is analyzed. Then, an optimal control point selection method based on the minimum resultant force principle is proposed to achieve precise manipulation of the breast target point. Concurrently, a breast soft tissue manipulation framework incorporating a Model Reference Adaptive Control (MRAC) system is developed to enhance operational accuracy. A dynamic model of breast soft tissue is developed by using the manipulative force–displacement data obtained during the process of manipulating breast soft tissue with mechanical fingers to realize the manipulative force control of breast tissue. Finally, through simulation and experiments on breast target-point manipulation tasks, the results show that this robotic system can achieve spatial control of breast positioning at arbitrary points. Meanwhile, the robotic system proposed in this study demonstrates high-precision control with an accuracy of approximately 1.158 mm (standard deviation: 0.119 mm), fulfilling the requirements for clinical interventional surgery in target point manipulation. Full article

(This article belongs to the Section Medical Robotics and Service Robotics)

► Show Figures

Figure 1

16 pages, 605 KiB

Open AccessArticle

Kriging-Variance-Informed Multi-Robot Path Planning and Task Allocation for Efficient Mapping of Soil Properties

by Laurence Roberts-Elliott, Gautham P. Das and Grzegorz Cielniak

Robotics 2025, 14(6), 77; https://doi.org/10.3390/robotics14060077 - 31 May 2025

Abstract

One of the most commonly performed environmental explorations is soil sampling to identify soil properties of agricultural fields, which can inform the farmer about the variable rate treatment of fertilisers in precision agriculture. However, traditional manual methods are slow, costly, and yield low [...] Read more.

One of the most commonly performed environmental explorations is soil sampling to identify soil properties of agricultural fields, which can inform the farmer about the variable rate treatment of fertilisers in precision agriculture. However, traditional manual methods are slow, costly, and yield low spatial resolution. Deploying multiple robots with proximal sensors can address this challenge by parallelising the sampling process. Yet, multi-robot soil sampling is under-explored in the literature. This paper proposes an auction-based multi-robot task allocation that efficiently coordinates the sampling, coupled with a dynamic sampling strategy informed by Kriging variance from interpolation. This strategy aims to reduce the number of samples needed for accurate mapping by exploring and sampling areas that maximise information gained per sample. The key innovative contributions include (1) a novel Distance Over Variance (DOV) bid calculation for auction-based multi-robot task allocation, which incentivises sampling in high-uncertainty, nearby areas; (2) integration of the DOV bid calculation into the cheapest insertion heuristic for task queuing; and (3) thresholding of newly created tasks at locations with low Kriging variance to drop those unlikely to offer significant information gain. The proposed methods were evaluated through comparative simulated experiments using historical soil compaction data. Evaluation trials demonstrate the suitability of the DOV bid calculation combined with task dropping, resulting in substantial improvements in key performance metrics, including mapping accuracy. While the experiments were conducted in simulation, the system is compatible with ROS and the ‘move_base’ action client to allow real-world deployment. The results from these simulations indicate that the Kriging-variance-informed approach can be applied to the exploration and mapping of other soil properties (e.g., pH, soil organic carbon, etc.) and environmental data. Full article

(This article belongs to the Special Issue Autonomous Robotics for Exploration)

► Show Figures

Figure 1

28 pages, 6914 KiB

Open AccessArticle

Guided Reinforcement Learning with Twin Delayed Deep Deterministic Policy Gradient for a Rotary Flexible-Link System

by Carlos Saldaña Enderica, José Ramon Llata and Carlos Torre-Ferrero

Robotics 2025, 14(6), 76; https://doi.org/10.3390/robotics14060076 - 31 May 2025

Abstract

This study proposes a robust methodology for vibration suppression and trajectory tracking in rotary flexible-link systems by leveraging guided reinforcement learning (GRL). The approach integrates the twin delayed deep deterministic policy gradient (TD3) algorithm with a linear quadratic regulator (LQR) acting as a [...] Read more.

This study proposes a robust methodology for vibration suppression and trajectory tracking in rotary flexible-link systems by leveraging guided reinforcement learning (GRL). The approach integrates the twin delayed deep deterministic policy gradient (TD3) algorithm with a linear quadratic regulator (LQR) acting as a guiding controller during training. Flexible-link mechanisms common in advanced robotics and aerospace systems exhibit oscillatory behavior that complicates precise control. To address this, the system is first identified using experimental input-output data from a Quanser^® virtual plant, generating an accurate state-space representation suitable for simulation-based policy learning. The hybrid control strategy enhances sample efficiency and accelerates convergence by incorporating LQR-generated trajectories during TD3 training. Internally, the TD3 agent benefits from architectural features such as twin critics, delayed policy updates, and target action smoothing, which collectively improve learning stability and reduce overestimation bias. Comparative results show that the guided TD3 controller achieves superior performance in terms of vibration damping, transient response, and robustness, when compared to conventional LQR, fuzzy logic, neural networks, and GA-LQR approaches. Although the controller was validated using a high-fidelity digital twin, it has not yet been deployed on the physical plant. Future work will focus on real-time implementation and structural robustness testing under parameter uncertainty. Overall, this research demonstrates that guided reinforcement learning can yield stable and interpretable policies that comply with classical control criteria, offering a scalable and generalizable framework for intelligent control of flexible mechanical systems. Full article

(This article belongs to the Section Industrial Robots and Automation)

► Show Figures

Figure 1

44 pages, 3893 KiB

Open AccessSystematic Review

Task Scheduling with Mobile Robots—A Systematic Literature Review

by Catarina Rema, Pedro Costa, Manuel Silva and Eduardo J. Solteiro Pires

Robotics 2025, 14(6), 75; https://doi.org/10.3390/robotics14060075 - 30 May 2025

Abstract

The advent of Industry 4.0, driven by automation and real-time data analysis, offers significant opportunities to revolutionize manufacturing, with mobile robots playing a central role in boosting productivity. In smart job shops, scheduling tasks involves not only assigning work to machines but also [...] Read more.

The advent of Industry 4.0, driven by automation and real-time data analysis, offers significant opportunities to revolutionize manufacturing, with mobile robots playing a central role in boosting productivity. In smart job shops, scheduling tasks involves not only assigning work to machines but also managing robot allocation and travel times, thus extending traditional problems like the Job Shop Scheduling Problem (JSSP) and Traveling Salesman Problem (TSP). Common solution methods include heuristics, metaheuristics, and hybrid methods. However, due to the complexity of these problems, existing models often struggle to provide efficient optimal solutions. Machine learning, particularly reinforcement learning (RL), presents a promising approach by learning from environmental interactions, offering effective solutions for task scheduling. This systematic literature review analyzes 71 papers published between 2014 and 2024, critically evaluating the current state of the art of task scheduling with mobile robots. The review identifies the increasing use of machine learning techniques and hybrid approaches to address more complex scenarios, thanks to their adaptability. Despite these advancements, challenges remain, including the integration of path planning and obstacle avoidance in the task scheduling problem, which is crucial for making these solutions stable and reliable for real-world applications and scaling for larger fleets of robots. Full article

(This article belongs to the Section Industrial Robots and Automation)

► Show Figures

Figure 1

37 pages, 13864 KiB

Open AccessArticle

LSTM-Enhanced Deep Reinforcement Learning for Robust Trajectory Tracking Control of Skid-Steer Mobile Robots Under Terra-Mechanical Constraints

by Jose Manuel Alcayaga, Oswaldo Anibal Menéndez, Miguel Attilio Torres-Torriti, Juan Pablo Vásconez, Tito Arévalo-Ramirez and Alvaro Javier Prado Romo

Robotics 2025, 14(6), 74; https://doi.org/10.3390/robotics14060074 - 29 May 2025

Abstract

Autonomous navigation in mining environments is challenged by complex wheel–terrain interaction, traction losses caused by slip dynamics, and sensor limitations. This paper investigates the effectiveness of Deep Reinforcement Learning (DRL) techniques for the trajectory tracking control of skid-steer mobile robots operating under terra-mechanical [...] Read more.

Autonomous navigation in mining environments is challenged by complex wheel–terrain interaction, traction losses caused by slip dynamics, and sensor limitations. This paper investigates the effectiveness of Deep Reinforcement Learning (DRL) techniques for the trajectory tracking control of skid-steer mobile robots operating under terra-mechanical constraints. Four state-of-the-art DRL algorithms, i.e., Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), Twin Delayed DDPG (TD3), and Soft Actor–Critic (SAC), are selected to evaluate their ability to generate stable and adaptive control policies under varying environmental conditions. To address the inherent partial observability in real-world navigation, this study presents an original approach that integrates Long Short-Term Memory (LSTM) networks into DRL-based controllers. This allows control agents to retain and leverage temporal dependencies to infer unobservable system states. The developed agents were trained and tested in simulations and then assessed in field experiments under uneven terrain and dynamic model parameter changes that lead to traction losses in mining environments, targeting various trajectory tracking tasks, including lemniscate and squared-type reference trajectories. This contribution strengthens the robustness and adaptability of DRL agents by enabling better generalization of learned policies compared with their baseline counterparts, while also significantly improving trajectory tracking performance. In particular, LSTM-based controllers achieved reductions in tracking errors of 10%, 74%, 21%, and 37% for DDPG-LSTM, PPO-LSTM, TD3-LSTM, and SAC-LSTM, respectively, compared with their non-recurrent counterparts. Furthermore, DDPG-LSTM and TD3-LSTM reduced their control effort through the total variation in control input by 15% and 20% compared with their respective baseline controllers, respectively. Findings from this work provide valuable insights into the role of memory-augmented reinforcement learning for robust motion control in unstructured and high-uncertainty environments. Full article

(This article belongs to the Section Intelligent Robots and Mechatronics)

► Show Figures

Figure 1

17 pages, 2621 KiB

Open AccessArticle

May I Assist You?—Exploring the Impact of Telepresence System Design on the Social Perception of Remote Assistants in Collaborative Assembly Tasks

by Jennifer Brade, Sarah Mandl, Franziska Klimant, Anja Strobel, Philipp Klimant and Martin Dix

Robotics 2025, 14(6), 73; https://doi.org/10.3390/robotics14060073 - 28 May 2025

Abstract

Remote support in general is a method that saves time and resources. A relatively new and promising technology for remote support that combines video conferencing and physical mobility is that of telepresence systems. The remote assistant, that is, the user of said technology, [...] Read more.

Remote support in general is a method that saves time and resources. A relatively new and promising technology for remote support that combines video conferencing and physical mobility is that of telepresence systems. The remote assistant, that is, the user of said technology, gains both presence and maneuverability in the distant location. As telepresence systems vary greatly in their design, the question arises as to whether the design influences the perception of the remote assistant. Unlike pure design studies, the present work focuses not only on the design and evaluation of the telepresence system itself, but especially on its perception during a collaborative task involving a human partner visible through the telepresence system. This paper presents two studies in which participants performed an assembly task under the guidance of a remote assistant. The remote assistant was visible through differently designed telepresence systems that were evaluated in terms of social perception and trustworthiness. Four telepresence systems were evaluated in study 1 (N = 32) and five different systems in study 2 (N = 34). The results indicated that similarly designed systems showed only marginal differences, but a system that was designed to transport additional loads and was therefore less agile and rather bulky was rated significantly less positively regarding competence than the other systems. It is particularly noteworthy that it was not the height of the communication medium that was decisive for the rating, but above all, the agility and mobility of the system. These results provide evidence that the design of a telepresence system can influence the social perception of the remote assistant and therefore has implications for the acceptance and use of telepresence systems. Full article

(This article belongs to the Special Issue Extended Reality and AI Empowered Robots)

► Show Figures

Figure 1

20 pages, 579 KiB

Open AccessArticle

Model-Based Predictive Control for Position and Orientation Tracking in a Multilayer Architecture for a Three-Wheeled Omnidirectional Mobile Robot

by Elena Villalba-Aguilera, Joaquim Blesa and Pere Ponsa

Robotics 2025, 14(6), 72; https://doi.org/10.3390/robotics14060072 - 28 May 2025

Abstract

This paper presents the design and implementation of a Model-based Predictive Control (MPC) strategy integrated within a modular multilayer architecture for a three-wheeled omnidirectional mobile robot, the Robotino 4 from Festo. The implemented architecture is organized into three hierarchical layers to support modularity [...] Read more.

This paper presents the design and implementation of a Model-based Predictive Control (MPC) strategy integrated within a modular multilayer architecture for a three-wheeled omnidirectional mobile robot, the Robotino 4 from Festo. The implemented architecture is organized into three hierarchical layers to support modularity and system scalability. The upper layer is responsible for trajectory planning. This planned trajectory is forwarded to the intermediate layer, where the MPC computes the optimal velocity commands to follow the reference path, taking into account the kinematic model and actuator constraints of the robot. Finally, these velocity commands are processed by the lower layer, which uses three independent PID controllers to regulate the individual wheel speeds. To evaluate the proposed control scheme, it was implemented in MATLAB R2024a using a lemniscate trajectory as the reference. The MPC problem was formulated as a quadratic optimization problem that considered the three states: the global position coordinates and orientation angle. The simulation included state estimation errors and motor dynamics, which were experimentally identified to closely match real-world behavior. The simulation and experimental results demonstrate the capability of the MPC to track the lemniscate trajectory efficiently. Notably, the close agreement between the simulated and experimental results validated the fidelity of the simulation model. In a real-world scenario, the MPC controller enabled simultaneous regulation of both the position and orientation, which offered a greater performance compared with approaches that assume a constant orientation. Full article

(This article belongs to the Section Sensors and Control in Robotics)

► Show Figures

Graphical abstract

17 pages, 1922 KiB

Open AccessArticle

Enhancing Visual–Inertial Odometry Robustness and Accuracy in Challenging Environments

by Alessandro Minervini, Adrian Carrio and Giorgio Guglieri

Robotics 2025, 14(6), 71; https://doi.org/10.3390/robotics14060071 - 27 May 2025

Abstract

Visual–Inertial Odometry (VIO) algorithms are widely adopted for autonomous drone navigation in GNSS-denied environments. However, conventional monocular and stereo VIO setups often lack robustness under challenging environmental conditions or during aggressive maneuvers, due to the sensitivity of visual information to lighting, texture, and [...] Read more.

Visual–Inertial Odometry (VIO) algorithms are widely adopted for autonomous drone navigation in GNSS-denied environments. However, conventional monocular and stereo VIO setups often lack robustness under challenging environmental conditions or during aggressive maneuvers, due to the sensitivity of visual information to lighting, texture, and motion blur. In this work, we enhance an existing open-source VIO algorithm to improve both the robustness and accuracy of the pose estimation. First, we integrate an IMU-based motion prediction module to improve feature tracking across frames, particularly during high-speed movements. Second, we extend the algorithm to support a multi-camera setup, which significantly improves tracking performance in low-texture environments. Finally, to reduce the computational complexity, we introduce an adaptive feature selection strategy that dynamically adjusts the detection thresholds according to the number of detected features. Experimental results validate the proposed approaches, demonstrating notable improvements in both accuracy and robustness across a range of challenging scenarios. Full article

(This article belongs to the Section Sensors and Control in Robotics)

► Show Figures

Figure 1

20 pages, 10201 KiB

Open AccessArticle

On First-Principle Robot Building in Undergraduate Robotics Education in the Robotic System Levels Model

by Bryan Van Scoy, Peter Jamieson and Veena Chidurala

Robotics 2025, 14(6), 70; https://doi.org/10.3390/robotics14060070 - 27 May 2025

Abstract

Robotics has widespread applications throughout industrial automation, autonomous vehicles, agriculture, and more. For these reasons, undergraduate education has begun to focus on preparing engineering students to directly contribute to the design and use of such systems. However, robotics is inherently multi-disciplinary and requires [...] Read more.

Robotics has widespread applications throughout industrial automation, autonomous vehicles, agriculture, and more. For these reasons, undergraduate education has begun to focus on preparing engineering students to directly contribute to the design and use of such systems. However, robotics is inherently multi-disciplinary and requires knowledge of controls and automation, embedded systems, sensors, signal processing, algorithms, and artificial intelligence. This makes training the future robotics workforce a challenge. In this paper, we evaluate our experiences with project-based learning approaches to teaching robotics at the undergraduate level at Miami University. Specifically, we analyze three consecutive years of capstone design projects on increasingly complex robotics design problems for multi-robot systems. We also evaluate the laboratories taught in our course “ECE 314: Elements of Robotics”. We have chosen these four experiences since they focus on the use of “cheap” first-principled robots, meaning that these robots sit on the fringe of embedded system design in that much of the student time is spent on working with a micro-controller interfacing with simple and cheap actuators and sensors. To contextualize our results, we propose the Robotic System Levels (RSL) model as a structured way to understand the levels of abstraction in robotic systems. Our main conclusion from these case studies is that, in each experience, students are exposed primarily to a subset of levels in the RSL model. Therefore, the curriculum should be designed to emphasize levels that align with educational objectives and the skills required by local industries. Full article

(This article belongs to the Section Educational Robotics)

► Show Figures

Figure 1

Journal Description

Robotics

Latest Articles

Journal Menu

Journal Browser

Highly Accessed Articles

Latest Books

E-Mail Alert

News

Topics

Conferences

Special Issues

Further Information

Guidelines

MDPI Initiatives

Follow MDPI