In Human–Robot Interaction(HRI),generating robot trajectories that accurately reflect user intentions while ensuring physical realism remains challenging,especially in unstructured environments.In this study,we devel...In Human–Robot Interaction(HRI),generating robot trajectories that accurately reflect user intentions while ensuring physical realism remains challenging,especially in unstructured environments.In this study,we develop a multimodal framework that integrates symbolic task reasoning with continuous trajectory generation.The approach employs transformer models and adversarial training to map high-level intent to robotic motion.Information from multiple data sources,such as voice traits,hand and body keypoints,visual observations,and recorded paths,is integrated simultaneously.These signals are mapped into a shared representation that supports interpretable reasoning while enabling smooth and realistic motion generation.Based on this design,two different learning strategies are investigated.In the first step,grammar-constrained Linear Temporal Logic(LTL)expressions are created from multimodal human inputs.These expressions are subsequently decoded into robot trajectories.The second method generates trajectories directly from symbolic intent and linguistic data,bypassing an intermediate logical representation.Transformer encoders combine multiple types of information,and autoregressive transformer decoders generate motion sequences.Adding smoothness and speed limits during training increases the likelihood of physical feasibility.To improve the realism and stability of the generated trajectories during training,an adversarial discriminator is also included to guide them toward the distribution of actual robot motion.Tests on the NATSGLD dataset indicate that the complete system exhibits stable training behaviour and performance.In normalised coordinates,the logic-based pipeline has an Average Displacement Error(ADE)of 0.040 and a Final Displacement Error(FDE)of 0.036.The adversarial generator makes substantially more progress,reducing ADE to 0.021 and FDE to 0.018.Visual examination confirms that the generated trajectories closely align with observed motion patterns while preserving smooth temporal dynamics.展开更多
Two new heuristic models are developed for motion planning of point robots in known environments.The first model is a combination of an improved particle swarm optimization (PSO) algorithm used as a global planner and...Two new heuristic models are developed for motion planning of point robots in known environments.The first model is a combination of an improved particle swarm optimization (PSO) algorithm used as a global planner and the probabilistic roadmap (PRM) method acting as a local obstacle avoidance planner.For the PSO component,new improvements are proposed in initial particle generation,the weighting mechanism,and position-and velocity-updating processes.Moreover,two objective functions which aim to minimize the path length and oscillations,govern the robot’s movements towards its goal.The PSO and PRM components are further intertwined by incorporating the best PSO particles into the randomly generated PRM.The second model combines a genetic algorithm component with the PRM method.In this model,new specific selection,mutation,and crossover operators are designed to evolve the population of discrete particles located in continuous space.Thorough comparisons of the developed models with each other,and against the standard PRM method,show the advantages of the PSO method.展开更多
This paper explores the realization of robotic motion planning, especially Findpath problem, which is a basic motion planning problem that arises in the development of robotics. Findpath means: Give the initial and de...This paper explores the realization of robotic motion planning, especially Findpath problem, which is a basic motion planning problem that arises in the development of robotics. Findpath means: Give the initial and desired final configurations of a robotic arm in 3-dimensionnl space, and give descriptions of the obstacles in the space, determine whether there is a continuous collision-free motion of the robotic arm from one configure- tion to the other and find such a motion if it exists. There are several branches of approach in motion planning area, but in reality the important things are feasibility, efficiency and accuracy of the method. In this paper ac- cording to the concepts of Configuration Space (C-Space) and Rotation Mapping Graph (RMG) discussed in [1], a topological method named Dimension Reduction Method (DRM) for investigating the connectivity of the RMG (or the topologic structure of the RMG )is presented by using topologic technique. Based on this ap- proach the Findpath problem is thus transformed to that of finding a connected way in a finite Characteristic Network (CN). The method has shown great potentiality in practice. Here a simulation system is designed to embody DRM and it is in sight that DRM can he adopted in the first overall planning of real robot sys- tem in the near future.展开更多
Obstacle avoidance is quite an important issue in the field of legged robotic applications, such as rescuing and detecting in complicated environment. Most related researchers focused on the legged robot’s gait gener...Obstacle avoidance is quite an important issue in the field of legged robotic applications, such as rescuing and detecting in complicated environment. Most related researchers focused on the legged robot’s gait generation after ssuming that obstacles have been detected and the walking path has been given. In this paper we propose and validate a novel obstacle avoidance framework for a six-legged walking robot Hexapod-III in unknown environment. Throughout the paper we highlight three themes: (1) The terrain map modeling and the obstacle detection; (2) the obstacle avoidance path planning method; (3) motion planning for the legged robot. Concretely, a novel geometric feature grid map (GFGM) is proposed to describe the terrain. Based on the GFGM, the obstacle detection algorithm is presented. Then the concepts of virtual obstacles and safe conversion pose are introduced. Virtual obstacles restrict the robot to walk on the detection terrain. A safe path based on Bezier curves, passing through safe conversion poses, is obtained by minimizing a penalty function taking into account the path length subjected to obstacle avoidance. Thirdly, motion planning for the legged robot to walk along the generated path is discussed in detail. At last, we apply the proposed framework to the Hexapod-III robot. The experimental result shows that our methodology allows the robot to walk safely without encountering with any obstacles in unknown environment.展开更多
基金The authors extend their appreciation to Prince Sattam bin Abdulaziz University for funding this research work through the project number(PSAU/2024/01/32082).
文摘In Human–Robot Interaction(HRI),generating robot trajectories that accurately reflect user intentions while ensuring physical realism remains challenging,especially in unstructured environments.In this study,we develop a multimodal framework that integrates symbolic task reasoning with continuous trajectory generation.The approach employs transformer models and adversarial training to map high-level intent to robotic motion.Information from multiple data sources,such as voice traits,hand and body keypoints,visual observations,and recorded paths,is integrated simultaneously.These signals are mapped into a shared representation that supports interpretable reasoning while enabling smooth and realistic motion generation.Based on this design,two different learning strategies are investigated.In the first step,grammar-constrained Linear Temporal Logic(LTL)expressions are created from multimodal human inputs.These expressions are subsequently decoded into robot trajectories.The second method generates trajectories directly from symbolic intent and linguistic data,bypassing an intermediate logical representation.Transformer encoders combine multiple types of information,and autoregressive transformer decoders generate motion sequences.Adding smoothness and speed limits during training increases the likelihood of physical feasibility.To improve the realism and stability of the generated trajectories during training,an adversarial discriminator is also included to guide them toward the distribution of actual robot motion.Tests on the NATSGLD dataset indicate that the complete system exhibits stable training behaviour and performance.In normalised coordinates,the logic-based pipeline has an Average Displacement Error(ADE)of 0.040 and a Final Displacement Error(FDE)of 0.036.The adversarial generator makes substantially more progress,reducing ADE to 0.021 and FDE to 0.018.Visual examination confirms that the generated trajectories closely align with observed motion patterns while preserving smooth temporal dynamics.
文摘Two new heuristic models are developed for motion planning of point robots in known environments.The first model is a combination of an improved particle swarm optimization (PSO) algorithm used as a global planner and the probabilistic roadmap (PRM) method acting as a local obstacle avoidance planner.For the PSO component,new improvements are proposed in initial particle generation,the weighting mechanism,and position-and velocity-updating processes.Moreover,two objective functions which aim to minimize the path length and oscillations,govern the robot’s movements towards its goal.The PSO and PRM components are further intertwined by incorporating the best PSO particles into the randomly generated PRM.The second model combines a genetic algorithm component with the PRM method.In this model,new specific selection,mutation,and crossover operators are designed to evolve the population of discrete particles located in continuous space.Thorough comparisons of the developed models with each other,and against the standard PRM method,show the advantages of the PSO method.
文摘This paper explores the realization of robotic motion planning, especially Findpath problem, which is a basic motion planning problem that arises in the development of robotics. Findpath means: Give the initial and desired final configurations of a robotic arm in 3-dimensionnl space, and give descriptions of the obstacles in the space, determine whether there is a continuous collision-free motion of the robotic arm from one configure- tion to the other and find such a motion if it exists. There are several branches of approach in motion planning area, but in reality the important things are feasibility, efficiency and accuracy of the method. In this paper ac- cording to the concepts of Configuration Space (C-Space) and Rotation Mapping Graph (RMG) discussed in [1], a topological method named Dimension Reduction Method (DRM) for investigating the connectivity of the RMG (or the topologic structure of the RMG )is presented by using topologic technique. Based on this ap- proach the Findpath problem is thus transformed to that of finding a connected way in a finite Characteristic Network (CN). The method has shown great potentiality in practice. Here a simulation system is designed to embody DRM and it is in sight that DRM can he adopted in the first overall planning of real robot sys- tem in the near future.
基金supported by the National Basic Research Program of China (Grant No. 2013CB035501)
文摘Obstacle avoidance is quite an important issue in the field of legged robotic applications, such as rescuing and detecting in complicated environment. Most related researchers focused on the legged robot’s gait generation after ssuming that obstacles have been detected and the walking path has been given. In this paper we propose and validate a novel obstacle avoidance framework for a six-legged walking robot Hexapod-III in unknown environment. Throughout the paper we highlight three themes: (1) The terrain map modeling and the obstacle detection; (2) the obstacle avoidance path planning method; (3) motion planning for the legged robot. Concretely, a novel geometric feature grid map (GFGM) is proposed to describe the terrain. Based on the GFGM, the obstacle detection algorithm is presented. Then the concepts of virtual obstacles and safe conversion pose are introduced. Virtual obstacles restrict the robot to walk on the detection terrain. A safe path based on Bezier curves, passing through safe conversion poses, is obtained by minimizing a penalty function taking into account the path length subjected to obstacle avoidance. Thirdly, motion planning for the legged robot to walk along the generated path is discussed in detail. At last, we apply the proposed framework to the Hexapod-III robot. The experimental result shows that our methodology allows the robot to walk safely without encountering with any obstacles in unknown environment.
