Robotics and Autonomous Systems最新文献

{"title":"On-policy and Off-policy Q-learning algorithms with policy iteration for two-wheeled inverted pendulum systems","authors":"Ba Quoc Anh Nguyen ,&nbsp;Ngoc Trung Dang ,&nbsp;Thanh Tung Le ,&nbsp;Phuong Nam Dao","doi":"10.1016/j.robot.2025.105111","DOIUrl":"10.1016/j.robot.2025.105111","url":null,"abstract":"<div><div>This article delves into the investigation of On-policy and Off-policy Q-learning algorithms for controlling two-wheeled inverted pendulum (TWIP) robots in situations where knowledge about the dynamic system is uncertain. Both on-policy and off-policy Q-learning algorithms ensure optimal and model-free control by employing a data collection approach without the knowledge of model. The On-policy algorithm performs real-time data collection, continuously gathering data and iteratively calculating a new control policy until it converges to the optimal value. In contrast, the Off-policy algorithm collects data only once and applies it to the system after completing the learning process. To enhance computational efficiency and minimize the amount of data required, the TWIP system is divided into two Sub-systems. These Sub-systems consist of smaller system matrices that can be controlled independently. This division reduces the data collection burden and accelerates the calculation speed of the algorithms. The utilization of Off-policy techniques proves to be advantageous in developing algorithms with data efficiency and achieving higher accuracy. The influence of probing noise on the Q-function is comprehensively considered in both proposed algorithms. By utilizing a single data set and eliminating the influence of noise, the Off-policy techniques enhance algorithm performance. Finally, the article presents simulation results of the TWIP system to validate the effectiveness of the two proposed control schemes.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105111"},"PeriodicalIF":4.3,"publicationDate":"2025-06-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144523056","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"HighwayLLM: Decision-making and navigation in highway driving with RL-informed language model","authors":"Mustafa Yildirim,&nbsp;Barkin Dagda,&nbsp;Vinal Asodia,&nbsp;Saber Fallah","doi":"10.1016/j.robot.2025.105114","DOIUrl":"10.1016/j.robot.2025.105114","url":null,"abstract":"<div><div>Autonomous driving is a complex task which requires advanced decision making and control algorithms. Understanding the rationale behind the autonomous vehicles’ decision is crucial to ensure their safe and effective operation on highway driving. This study presents a novel approach, <em>HighwayLLM</em>, which harnesses the reasoning capabilities of large language models (LLMs) to predict the future waypoints for ego-vehicle’s navigation. Our approach also utilizes a pre-trained Reinforcement Learning (RL) model to serve as a high-level planner, making decisions on appropriate meta-level actions. The HighwayLLM combines the output from the RL model and the current state information to make safe, collision-free, and explainable predictions for the next states, thereby constructing a trajectory for the ego-vehicle. Subsequently, a PID-based controller guides the vehicle to the waypoints predicted by the LLM agent. This integration of LLM with RL and PID enhances the decision-making process, provides interpretability for highway autonomous driving and reduces the number of collisions compared to the baseline method.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105114"},"PeriodicalIF":4.3,"publicationDate":"2025-06-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144523181","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Challenges: Strong impact operation robots—motion analysis, transient measurement, and vibration suppression strategy","authors":"Tong Mou ,&nbsp;Xianzhong Chen ,&nbsp;Qingwen Hou ,&nbsp;Jiahao Hua ,&nbsp;Zeping Hu ,&nbsp;Chaoda Xie ,&nbsp;Haisheng Zhong ,&nbsp;Zhikun Qi","doi":"10.1016/j.robot.2025.105112","DOIUrl":"10.1016/j.robot.2025.105112","url":null,"abstract":"<div><div>As industry tends to be much intelligence-oriented, the performance of industrial robots faces increasingly a higher demand. Strong Impact Operation Robots (SIORs), additionally with high efficiency and durability, have been widely applied in industry. However, with a feature of high-speed, non-linear and discontinuous impact and heavy load, SIORs inevitably trigger more complex and intense vibrations during the motion process, which restricts the improvement of the operational accuracy, stability and efficiency of such robots. This paper provides an overview of researches on vibration suppression of SIORs, and divides them into three categories, motion analysis and modeling, vibration sensing of impact transients, and design of vibration suppression strategies for impact transients. Taking the metallurgical furnace front operation robot as an example and combining it with real industrial scenarios, this paper points out the major challenges for SIORs in each category, setting up an accurate dynamic model of impact mechanism in a variable temperature environment, so as to realize vibration transient measurements under high-speed, non-linear and discontinuous impact, and cultivate active and passive vibration suppression techniques and model-data-driven strategies. This paper systematically investigates the complex vibration problem in SIORs, of which the findings are significant for enhancing the accuracy, stability and efficiency of SIORs, providing a solid foundation for future research on the complex vibration issues of robots.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105112"},"PeriodicalIF":4.3,"publicationDate":"2025-06-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144513654","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A system for docking robotic wheelchair to partially visible table of unknown pose using human input and robotic wheelchair motion","authors":"Neha Priyadarshini Garg,&nbsp;Marcus Leong,&nbsp;Manoj Ramanathan,&nbsp;Wee-Ching Pang,&nbsp;Lei Li,&nbsp;Wei Tech Ang","doi":"10.1016/j.robot.2025.105087","DOIUrl":"10.1016/j.robot.2025.105087","url":null,"abstract":"<div><div>Existing autonomous table docking approaches require large part of table to be visible in order to compute good docking pose for a table with unknown pose. However, in real world settings like cluttered offices and food courts, users often need to initiate docking when the table is only partially visible. This work focuses on the development of a table docking system that enables a wheelchair user to dock to a table with unknown pose, even when the table is partially visible by using human input and the wheelchair motion. An intuitive point-and-click interface is implemented to allow user to initiate docking, by simply clicking on table in the RGB image of the scene. Docking pose is calculated by extracting table edge from the point cloud using the information provided by the user’s click and the streaming RGBD camera images of the scene as the wheelchair moves towards the table. Based on our experiments, accurate table docking can be achieved even when the table is only 20% visible using our approach while the baseline system requires at least 70% of the table to be visible. This makes our system applicable in realistic scenarios. During evaluation with human subjects, all the participants preferred our system over the baseline due to ease of use. Quantitatively, our system more than halved the effort required to initiate autonomous docking as compared to the baseline.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105087"},"PeriodicalIF":4.3,"publicationDate":"2025-06-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144491092","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Coverage Trajectory Planning Problem on 3D Terrains with safety constraints for automated lawn mower: Exact and heuristic approaches","authors":"Hang Zhou,&nbsp;Peng Zhang,&nbsp;Zhaohui Liang,&nbsp;Hangyu Li,&nbsp;Xiaopeng Li","doi":"10.1016/j.robot.2025.105109","DOIUrl":"10.1016/j.robot.2025.105109","url":null,"abstract":"<div><div>Recent technological advancements in automation have attracted increased interest in automated lawn mowers. Developing a safe and efficient trajectory to cover entire terrains is crucial for autonomous mowing. Unlike the simplified indoor environments with flat surfaces commonly assumed in existing literature, lawn maintenance typically occurs in complex outdoor settings characterized by irregular terrains, including obstacles and slopes. These irregularities pose significant safety risks, such as the potential for the mower to tip over. This paper introduces the Coverage Trajectory Planning Problem on 3D Terrains (CTPP-3DT), which involves determining both the path and speed profile for an automated lawn mower to effectively cover a general 3D terrain with varying slopes. The objective of the CTPP-3DT is to minimize the completion time, including the time for turning, which satisfies safety constraints on the robot’s speed and acceleration on various slopes. To address this challenge, we first propose a Mixed-Integer Linear Programming (MILP) model based on a graph expansion method, suitable for solving small-scale instances. For larger instances, we develop a decomposition-based heuristic algorithm using Simulated Annealing. Extensive experiments conducted on benchmark instances demonstrate the effectiveness of our proposed MILP model and heuristic algorithm for small-size and large-size instances. The comparison between the optimized strategy and the conservative strategy highlights the necessity of incorporating safety constraints in trajectory planning, resulting in an average reduction of more than 40% in completion time. Furthermore, sensitivity analyses reveal that technological advancements in mowers, such as increasing the maximum speed and acceleration and reducing turning speed, can significantly reduce the overall completion time. Our code and data are available at <span><span>https://github.com/CATS-Lab/Mower-CTPP-3D</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105109"},"PeriodicalIF":4.3,"publicationDate":"2025-06-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144523055","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Adaptive coordinated impedance control for dual-arm robot symmetric bimanual tasks","authors":"Yang Zhang","doi":"10.1016/j.robot.2025.105110","DOIUrl":"10.1016/j.robot.2025.105110","url":null,"abstract":"<div><div>This paper proposes an adaptive coordinated impedance control scheme to address the challenges of internal force regulation and external force tracking in symmetric bimanual tasks performed by dual-arm robots. The scheme is developed within the hybrid impedance control framework, a specialized variant of the general impedance control approach. First, the contact force is decomposed using the dynamic model of dual-arm manipulation and integrated into the corresponding control loop for real-time impedance parameter adjustment. Then, based on the dynamic model of the closed-chain system, distinct impedance parameter adaptation strategies are designed for internal and external forces to meet their respective force control requirements during manipulation. By enabling real-time impedance parameter adjustments in response to contact force errors, the scheme effectively mitigates trajectory deviations caused by unknown paths and environmental factors, thereby improving dual-arm manipulation performance. Finally, the effectiveness of the adaptive coordinated control scheme is validated through a Matlab-Adams co-simulation on a dual-arm robot, demonstrating its capability in internal force limitation and external force tracking.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105110"},"PeriodicalIF":4.3,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144491090","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Time-varying control strategies for quadrotor dynamics with enhanced anti-windup capabilities","authors":"João Madeiras ,&nbsp;Carlos Cardeira ,&nbsp;Paulo Oliveira","doi":"10.1016/j.robot.2025.105107","DOIUrl":"10.1016/j.robot.2025.105107","url":null,"abstract":"<div><div>This paper proposes robust control strategies for position and attitude, incorporating anti-windup mechanisms to address actuator saturation and improve transient performance in quadrotors. The methodology features a robust, continuously differentiable, saturated position feedback law with anti-windup compensation, alongside a robust attitude tracking controller formulated in quaternion representation. This framework is particularly adept at handling constant biases in the moment of inertia and constant disturbances in both position and attitude closed-loop subsystems. A dynamic activation mechanism governs the anti-windup component, reducing overshoot and enhancing performance. Leveraging Matrosov’s Theorem, the proposed control architecture demonstrates robustness against bounded disturbances and ensures almost uniform global asymptotic stability of the full tracking error dynamics. Simulation results confirm the viability and effectiveness of the proposed approach.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105107"},"PeriodicalIF":4.3,"publicationDate":"2025-06-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144523057","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Learning multi-robot task allocation using capsule networks and attention mechanism","authors":"Steve Paul,&nbsp;Souma Chowdhury","doi":"10.1016/j.robot.2025.105085","DOIUrl":"10.1016/j.robot.2025.105085","url":null,"abstract":"<div><div>This paper presents a new graph reinforcement learning (RL) architecture to solve multi-robot task allocation (MRTA) problems without requiring any tedious heuristics. Multi-feature tasks are abstracted as nodes in an undirected graph in this case. The primary goal is to not only generalize across unseen problems of similar size but also scale to problems with much larger task spaces without retraining; which otherwise could be particularly expensive when simulating multi-robot operations. While drawing inspiration from the emerging paradigm in learning to solve combinatorial optimization (CO) problems, a new encoder–decoder architecture called Capsule Attention-based Mechanism or CAPAM is presented here to achieve this goal. More specifically, a novel choice of <em>encoder</em> is made in the form of graph capsule convolutional networks, which enables permutation invariant embeddings that capture the local and global structure of the task graph by using higher-order statistical moments of the vectors of node features. This encoded information is combined with a <em>context</em> component encoding mission and robot states, and processed through the <em>decoder</em> that computes the probability of selecting different available tasks by a robot. To train the CAPAM model, a policy-gradient method based on Proximal Policy Optimization is used. When evaluated over unseen scenarios, CAPAM demonstrates comparable task completion performance and faster decision-making compared to standard non-learning-based online MRTA methods. CAPAM demonstrates substantial gains in generalizability and (task) scalability in comparison to a popular approach for learning to solve CO problems (the pure attention mechanism) and preserves this performance advantage even under partial observation scenarios.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105085"},"PeriodicalIF":4.3,"publicationDate":"2025-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144491091","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Positioning of autonomous mobile robot using multi-lateration with pattern recognition and differential evolution","authors":"Tz-Yu Yang ,&nbsp;Brijesh Patel ,&nbsp;Po-Yan Chiu ,&nbsp;Chien-Ching Ma ,&nbsp;Chii-Rong Yang ,&nbsp;Chao-Lung Yang ,&nbsp;Po Ting Lin","doi":"10.1016/j.robot.2025.105113","DOIUrl":"10.1016/j.robot.2025.105113","url":null,"abstract":"<div><div>The extensive adoption of Autonomous Mobile Robots (AMR) in manufacturing, processing, and intelligent logistics has witnessed a remarkable increase, driven by the rapid growth of smart manufacturing and Industry 4.0. AMRs serve a dual role, facilitating both product handling and transportation. The precision of AMR positioning is of paramount importance. The prevalent approach to indoor positioning and navigation involves the use of cameras, optical Light Detection and Ranging (LiDAR) sensors. However, relying solely on LiDAR-based motion estimation for relative positioning can result in gradual displacement errors, impacting accuracy. This paper introduces a dual-positioning strategy to address this challenge, incorporating secondary localization methods to ensure precise spatial confirmation and task execution for a High Payload Autonomous Mobile Robot (HAMR). This proposed method integrates a RGB-D camera with the HAMR’s manipulator. It recognizes wall patterns (ArUco) and measures their distance from the HAMR, employing multi-lateration to calculate the HAMR’s position within the real-world coordinate system. This paper presents an indoor positioning method for HAMRs using ArUco code, enabling multi-lateration measurements within a 15 mm error. Differential Evolution (DE) is employed for motion analysis to solve inverse kinematics, enabling dynamic analysis of HAMRs with redundant degrees of freedom. This technique effectively compensates for positioning errors, significantly enhancing the AMR’s capabilities.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105113"},"PeriodicalIF":4.3,"publicationDate":"2025-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144491093","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Machine learning-based inverse kinematics scalability for prismatic tensegrity structural manipulators","authors":"Aidar Shakerimov ,&nbsp;Medetkhan Altymbek ,&nbsp;Koichi Koganezawa ,&nbsp;Azamat Yeshmukhametov","doi":"10.1016/j.robot.2025.105102","DOIUrl":"10.1016/j.robot.2025.105102","url":null,"abstract":"<div><div>Tensegrity structures are gaining attention due to their distinctive features that stem from wire-driven mechanisms and their highly redundant nature. These features include a lightweight framework, improved resistance to impacts, and ability to carry high payloads. Nonetheless, controlling these structures and understanding their movement remain complex challenges. Our research introduces a pioneering control strategy that utilizes some machine learning algorithms (linear regression, ridge regression, and neural network feedforward) to achieve inverse kinematics for prismatic tensegrity manipulators. This approach has been experimentally validated on two different structures, one with a triangular and the other with a quadrangular configuration, each forming a dual-layer setup. Our experimental results indicate that each of the presented algorithms facilitates the approximate inverse kinematics required for the control of the manipulators with average precision error of 2 cm.</div></div>","PeriodicalId":49592,"journal":{"name":"Robotics and Autonomous Systems","volume":"193 ","pages":"Article 105102"},"PeriodicalIF":4.3,"publicationDate":"2025-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144335990","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}