Design of an experimental platform for a dual-arm teleoperation mobile robot for embodied intelligence

With advances in new engineering education and the rapid evolution of embodied intelligence technologies, conventional robotic teaching platforms are increasingly constrained by limited functionality, strong software–hardware coupling, and poor extensibility for secondary development. Hence, they cannot effectively meet the emerging demands of practical teaching in embodied intelligence, multimodal interaction, and robotic manipulation. To address these challenges, this study designed and developed a dual-arm teleoperated mobile robot experimental platform. An integrated teaching framework encompassing “perception–decision–execution–interaction” was established using this platform to enhance students’ engineering practice and innovation capabilities in robotics and embodied intelligence, while supporting practical teaching and interdisciplinary integration in emerging engineering fields related to embodied intelligence.

The proposed platform has an integrated architecture comprising an omnidirectional mobile chassis, dual robotic arms, and an intelligent interaction module. Its software framework is based on the Robot Operating System (ROS). The platform integrates multiple functional modules, including chassis motion control, SLAM-based mapping and navigation, robotic arm teleoperation, and voice interaction. Multi-sensor data, including LiDAR, IMU, and odometry data, are fused to achieve environmental perception and autonomous motion control. The Cartographer algorithm is employed as the core framework for two-dimensional (2D) map construction for mapping and navigation. Path planning algorithms such as A*, DWA, and TEB are integrated for autonomous navigation and dynamic obstacle avoidance. The teleoperation subsystem employs a master–slave architecture based on socket communication to reproduce the motion trajectories of the dual robotic arms. The voice interaction module enables natural language–based robot control through speech recognition and keyword spotting technologies. Furthermore, a hierarchical experimental teaching system encompassing chassis drive control, ROS communication, mapping and navigation, teleoperation, and voice interaction was established. Practical training was conducted by integrating course-based instruction with exploratory experiments.

Experimental results demonstrated that the proposed platform operated stably with comprehensive functionality and exhibited robust capabilities in autonomous navigation, teleoperation, and intelligent interaction. The mobile chassis achieved smooth omnidirectional motion and rapid control response, and communication between the high-level and low-level controllers remained stable and reliable. The mapping and navigation subsystem successfully generates indoor 2D maps and supports autonomous navigation and dynamic obstacle avoidance under different path planning algorithms. In addition, it enables comparative experiments on various algorithms and parameter configurations. The dual-arm teleoperation subsystem achieves low-latency communication with stable motion synchronization and reproduction. The voice interaction subsystem successfully performs navigation control, operating mode switching, and predefined action triggering. The platform enables students to independently modify navigation parameters, interaction commands, and action sequences, thereby facilitating the transition from verification-based experiments to exploratory experiments. Thus, the platform effectively meets the requirements of robotic function validation and engineering practice-oriented training.

The proposed dual-arm teleoperated mobile robot experimental platform for embodied intelligence education integrates key technologies such as motion control, mapping and navigation, teleoperation, and voice interaction. Thus, a comprehensive experimental teaching system is established to cover low-level drivers, environmental perception, path planning, and intelligent interaction. The platform features high modularity, strong scalability, and cost-effectiveness, demonstrating excellent applicability and reproducibility in educational environments. It effectively cultivates students’ interdisciplinary competencies in embedded systems, robotics, and embodied intelligence. Moreover, it demonstrates strong research extensibility and can serve as an experimental foundation for future studies on imitation learning, reinforcement learning, and embodied intelligence driven by large models.