Advances in Robotics and Autonomous Systems: Transitioning from Automation to Intelligent Collaboration

1. Introduction

Since the Third Industrial Revolution, robotics and automated systems have become deeply integrated into industrial production and daily life, fundamentally reshaping production efficiency and product quality. With the ongoing advancement of the Fourth Industrial Revolution (Industry 4.0), a new paradigm is emerging—one defined by the seamless integration of physical systems and digital intelligence. In this evolving landscape, robots are no longer isolated machines executing predefined instructions; rather, they are transitioning into intelligent, autonomous agents capable of sustained and context-aware collaboration with humans [1].

Driven by rapid progress in Artificial Intelligence (AI) and the Internet of Things (IoT), the enabling technologies of robotics are advancing at unprecedented speed. The deep convergence of mechatronics, advanced materials, computer vision, and modern control theory provides the technological foundation for robots to transcend the constraints of structured factory environments and operate reliably in complex, dynamic, and unstructured settings. Synthesizing current research trajectories and engineering practices, the evolution of robotics and autonomous systems can be conceptually organized into four tightly coupled and mutually reinforcing layers.

Layer 1: Robotic Mechanisms and Morphology. At the foundational level, research is redefining robotic embodiment beyond traditional rigid-body architectures. Emerging paradigms include rigid–flexible hybrid systems [2], high-speed, high-precision parallel and hybrid mechanisms [3,4], bio-inspired and humanoid robots [5], and deformable soft robots [6]. While expanding application scenarios, these new forms introduce significant nonlinearity, high dimensionality, and difficult-to-model hysteretic effects, challenging traditional control theories and the entire perception-to-actuation pipeline.

Layer 2: Perception and Environmental Modeling. The deep integration of machine vision and machine learning has propelled robotic perception from geometric reconstruction toward semantic cognition [7]. Beyond achieving high-precision localization and robust trajectory estimation in dynamic, texture-poor, and low-illumination environments [8], current research increasingly emphasizes semantic mapping that integrates object attributes with contextual environmental relationships. Through multi-modal sensor fusion—including vision, LiDAR, thermal sensing, and tactile feedback—robots are developing all-weather, multi-dimensional perception frameworks capable of delivering reliable and interpretable state information to higher-level decision systems [9].

Layer 3: Control, Planning, and Digital Twins. For robots containing flexible actuators and structurally complex components that resist analytical modeling, data-driven paradigms and Reinforcement Learning (RL) have demonstrated strong adaptability and generalization capabilities [10]. Advanced control theories and optimization frameworks—such as convex optimization and Mixed-Integer Programming (MIP)—are increasingly deployed to address nonlinear motion planning and obstacle avoidance in high-degree-of-freedom systems [11]. Under hardware constraints and environmental uncertainty, planning architectures are evolving from deterministic path generation toward cognitively informed, task-level reasoning [12], while distributed coordination and Swarm Intelligence expand system scalability [13].

Moreover, the integration of Large Language Models (LLMs) introduces logical reasoning and long-horizon task planning capabilities, positioning robots as intelligent coordinators that bridge perception and execution [14]. In parallel, Digital Twin and Sim2Real technologies provide a critical infrastructure linking virtual training with physical deployment [15]. Through high-fidelity simulation and domain randomization, researchers are actively mitigating the “reality gap,” accelerating the translation of theoretical advances into robust engineering solutions [16].

Layer 4: Application Systems and Societal Scenarios. The ultimate validation of robotics lies in real-world deployment across industrial manufacturing [17], underwater and field operations [18], and healthcare and elderly care services [19]. At the system level, long-term reliability, safety assurance, interoperability, and social acceptance emerge as decisive indicators of technological maturity. Importantly, application-layer demands exert reverse constraints on the underlying mechanical, perceptual, and control layers, compelling continuous refinement toward greater intelligence, compliance, and trustworthiness.

Against this backdrop, this Special Issue, entitled “Advances in Robotics and Autonomous Systems”, was established as a platform for disseminating recent breakthroughs across these interconnected layers. The collected contributions span soft robot structural optimization, precision calibration of parallel mechanisms, perception and localization in mobile and underwater systems, and AI-driven societal demand analysis. Together, they reflect the dynamic evolution of modern robotics toward enhanced monitoring accuracy, control robustness, and operational reliability across diverse application domains.

2. An Overview of Published Articles

This Special Issue compiles eight high-quality research papers that concretely embody the four-layer technical framework outlined above. Through innovations in mechanism design, perception modeling, intelligent control, and application-oriented analytics, these works collectively illustrate how modern robotics advances from foundational hardware to system-level societal integration.

Layer 1: Mechanical Structure Innovation and Kinematic Optimization. Mechanical innovation remains the physical foundation of robotic intelligence. Regarding soft robotics, Khan et al. drew inspiration from reptilian scale friction to design a bio-inspired Kirigami anchoring structure, overcoming low locomotion efficiency in unstructured environments [20]. Through systematic geometric and actuation parameter optimization, this study demonstrated that specific triangular Kirigami configurations significantly enhance anisotropic friction properties. Validation demonstrated a stride efficiency of 63% and a crawling speed of approximately 47 cm/min, offering a compact and energy-efficient locomotion strategy for confined and complex environments.

Studying high-precision rigid mechanisms, Yu addressed the stringent calibration requirements of parallel platforms used in spacecraft on-orbit docking simulations [21]. An Improved Particle Swarm Optimization (PSO) algorithm was developed to convert complex kinematic parameter identification into a tractable high-dimensional nonlinear optimization problem. The experimental results showed that positioning error was reduced from 4.33 mm to 0.77 mm after calibration, accompanied by a significant improvement in attitude accuracy. This work highlights the increasing role of evolutionary optimization in enhancing structural precision and dynamic fidelity.

Layer 2: Precise Perception and Robust Localization. Perception serves as the cognitive gateway between the robot and its environment. To mitigate visual occlusion in complex object manipulation tasks, Hong et al. proposed a “See-Then-Grasp” strategy featuring a Two-Stage Active Reconstruction framework using a single robotic arm [22]. By autonomously optimizing viewpoints to eliminate blind spots, the system constructs high-fidelity 3D object models, significantly improving grasp success rates for previously unseen objects in unstructured settings. This work demonstrates the shift from passive perception to active, task-driven environmental reconstruction.

For indoor mobile robot localization in GPS-denied environments, Al-Hadithi and Pastor tackled the cost barrier of conventional LiDAR- or vision-based systems [23]. They designed a low-cost localization architecture based on ultrasonic beacons and differential Time-of-Flight (dToF) measurements. By refining differential ranging algorithms to suppress cumulative ultrasonic errors, the system achieved reliable navigation performance without expensive sensing infrastructure, offering a scalable pathway for large-scale deployment of logistics and service robots.

Layer 3: Reliable Control and High-Fidelity Simulation. Control and simulation technologies form the bridge between digital intelligence and physical execution. In fine manipulation, Park et al. proposed a Cartesian force-controlled Pushing Primitive to overcome the limitations of traditional pick-and-place strategies [24]. By introducing compliant force feedback, the system enabled precise in-hand adjustment without reliance on complex visual servoing. Experimental evaluations demonstrated placement errors controlled within 0.25 mm (X-axis) and 0.23 mm (Y-axis), providing a robust non-prehensile manipulation solution for precision assembly tasks.

In extreme environment simulation, Orjales et al. addressed the dynamics modeling challenges of open-frame Remotely Operated Vehicles (ROVs) by proposing a hybrid parameter identification method combining empirical measurement and Evolutionary Algorithms [25]. Using the BlueROV2 platform, they automatically optimized hydrodynamic coefficients to build a high-fidelity Digital Twin, effectively bridging the simulation-reality gap in complex underwater environments.

Similarly, Dong et al. focused on snow-removal field robots and established a systematic calibration methodology for snow particle contact parameters [26]. Using the Discrete Element Method (DEM) combined with the Hertz–Mindlin with JKR contact model and response surface analysis, they optimized contact parameters to keep the relative error of the simulated angle of repose within 0.32%, providing a reliable virtual verification benchmark for structural optimization.

Layer 4: Societal Application Analysis and System-Level Intelligence. At the societal systems level, robotics must align with demographic and ecological realities. Wen et al. applied machine learning techniques to conduct a large-scale quantitative analysis of community elderly care demands across major Chinese cities [27]. The resulting high-precision predictive model revealed nonlinear patterns in health conditions and lifestyle factors among elderly populations. Although centered on data analytics, the study provides essential socio-technical evidence for the functional design and deployment strategies of next-generation nursing robots and assistive systems, reinforcing the feedback loop between application demand and technological development.

3. Conclusions

Collectively, the contributions in this Special Issue delineate a coherent evolution of robotic systems toward greater intelligence, compliance, and societal integration across multiple system layers. Interpreting these works within a unified four-layer framework clarifies not only current technical achievements but also the structural trajectory and emerging bottlenecks shaping next-generation robotics.

Mechanisms. Robotic design is transitioning from a traditional emphasis on stiffness and load capacity toward paradigms centered on rigid–flexible fusion, variable stiffness, and intrinsically safe interaction. Parallel mechanisms, soft structures, and cable-driven architectures expand operational scale and adaptability, yet simultaneously introduce cross-scale coupling effects, nonlinear dynamics, and stringent demands on modeling and control fidelity.

Perception. Robotic perception is advancing from geometric reconstruction to semantic and physical understanding. The integration of multi-modal sensing and active perception strategies enables robots to capture not only spatial information but also physical constraints, contact states, and task-relevant semantics. Such interpretable environmental representations will form the cognitive substrate for higher-level reasoning and autonomous decision-making.

Control & Simulation. The convergence of physics-based modeling and data-driven learning is emerging as the dominant paradigm for complex robotic systems. High-fidelity Digital Twins, learning-augmented dynamic models, and robust Sim2Real transfer frameworks are essential to bridging the gap between virtual training and physical deployment, particularly in safety-critical and high-uncertainty environments.

Applications and Societal Integration. As robots increasingly operate in open, dynamic, and human-centered settings, system-level reliability, safety assurance, ethical governance, and social trust become defining criteria of technological maturity. These societal constraints exert reverse pressure on the underlying mechanical, perceptual, and control layers, driving continuous co-evolution toward more transparent, compliant, and trustworthy robotic systems.

We anticipate that this Special Issue will serve as a valuable reference for the systematic advancement of robotics and autonomous systems. By fostering dialogue across mechanism design, perception intelligence, control theory, and application ecosystems, it aims to accelerate the transition of robotic technologies from laboratory innovation to robust engineering deployment and meaningful societal impact.