bimanual coordination

Whole-body teleoperation often suffers from high operational complexity and cognitive load due to the need to simultaneously coordinate the head, arms, torso, and mobile base under kinematic and environmental constraints. This work proposes a coupled egocentric control method that naturally maps the operator’s head orientation (pitch/yaw) and hand poses to the robot’s torso height and base translation/rotation, enabling automatic coordination between the torso and base while allowing the user to focus solely on gaze direction and hand movements. Integrated with a workspace boundary-triggering mechanism, the approach eliminates the need for explicit command inputs. Evaluated on TIAGo in home-care tasks, the method significantly improves task efficiency over baseline approaches, reduces button presses and singular configurations, lowers cognitive load, and enhances usability and user preference.

This work addresses the challenge of coordinating a wheeled mobile base and dual arms during whole-body teleoperation of a mobile dual-arm robot, where operators struggle to simultaneously manage motion and avoid collisions. The authors propose an open-source teleoperation system that maps upper-limb motions to the robot arms while controlling the base via foot pedals. For the first time, lidar-driven haptic feedback is integrated into the foot pedals, complemented by bilateral force reflection and real-time manipulability visualization, enabling natural collision avoidance without explicit obstacle-avoidance controllers. Leveraging a low-cost lidar, multimodal force feedback, and the Action Chunking with Transformers (ACT) policy, the system demonstrates effective coordinated control in long-duration, high-precision tasks on a real dual-arm mobile platform, and shows that teleoperation-derived feedback signals can significantly enhance policy learning performance.

To address the challenge of driving high-degree-of-freedom (5–28 DoF) assistive devices using low-bandwidth, high-noise compensatory motions—particularly for users with motor impairments—this paper proposes a general human-robot collaborative control framework. Methodologically, it introduces the first unified “embodied extension” model for the full spectrum of assistive systems, integrating motion-intent decoding, compensatory-motion recognition, and adaptive inverse-dynamics mapping into a closed-loop control paradigm, validated via both virtual twin simulation and physical humanoid robot embodiment. The key contribution lies in cross-scale dynamic suppression of compensatory motions while preserving user-intent fidelity, enabling zero-shot generalization across devices of varying DOFs without retraining. Experiments demonstrate a 37% reduction in subjective fatigue and a 94.2% task-completion rate.

This study addresses the challenge of maintaining human postural balance during the use of general-purpose supernumerary limbs, which often disrupt stability while assisting with diverse tasks. To this end, the authors propose a novel three-layer hierarchical architecture grounded in human dynamics: a prediction layer that estimates real-time trunk and center-of-mass (CoM) states, a planning layer that generates optimal CoM trajectories to counteract disturbances, and a control layer that outputs limb actuation commands. This approach is the first to actively preserve human balance across general tasks without compromising functional versatility, overcoming the limitations of prior methods restricted to specific, static scenarios. Experimental results involving ten participants performing forward-bending and lateral-bending tasks demonstrate that the system significantly reduces postural instability, thereby enhancing both safety and comfort in human–robot collaboration.

To address the challenges of motion-manipulation coupling, weak environmental perception, and low intuitiveness in teleoperation of highly redundant embodied robots (e.g., CENTAURO), this paper proposes a wearable haptic human–robot interface inspired by the marionette metaphor. Our approach uniquely integrates multimodal haptic feedback—vibrotactile and force cues—deeply into a closed-loop motion mapping framework, synergizing virtual physical interaction modeling with a real-time motion–sensor co-mapping algorithm. This enables natural, full-platform kinematic mapping from operator limb motions to robot motion while concurrently rendering proprioceptive and environmental contact states. Experimental evaluation demonstrates that novice users achieve a 37% reduction in task completion time, a 52% decrease in collision misclassification rate, and a 41% reduction in NASA-TLX subjective workload—significantly enhancing operational efficiency, safety, and immersion.

In dual-arm human–robot collaborative manipulation, balancing ergonomic safety with force manipulation capability remains challenging. To address this, we propose a synergistic upper-limb pose optimization method integrating physical ergonomics and force manipulability. We formulate, for the first time, a coupled cost function based on joint-angle dependencies that jointly minimizes muscle activation load and maximizes force manipulability (i.e., minimizes the reciprocal of the force ellipsoid volume), guided by robot-end reference poses to steer the human into optimal collaborative postures. The method leverages a simplified skeletal model, analytical transformation modules, and a bimanual Model Predictive Impedance Controller (MPIC), enabling support for multiple grasp configurations and irregular object geometries. Experimental evaluations across multiple subjects and objects demonstrate a 23.6% reduction in target muscle activation, a 31.4% decrease in peak muscle load, and significant improvements in trajectory smoothness and force-tracking accuracy.

This work addresses the challenge of whole-body coordinated locomotion and manipulation control for quadrupedal robots equipped with robotic arms. We propose a reinforcement learning framework that integrates simulation and real-world data. Our method introduces two key innovations: (1) a trajectory library mechanism with adaptive curriculum sampling to enhance cross-task generalization; and (2) a trajectory-velocity prediction policy network that jointly models future states and enables zero-shot transfer. The framework supports both autonomous execution and teleoperation modes. A systematic ablation study is conducted in simulation, and multi-task zero-shot transfer is successfully demonstrated on a physical platform (Unitree Go2 quadruped + Airbot Arm). Experimental results show significant improvements in whole-body locomotion-manipulation coupling accuracy and task adaptability. This approach provides a scalable solution for multi-scale manipulation control in embodied agents.

Network-induced time delay compromises stability in robot-mediated dual-user haptic interaction systems. Method: We formulate a delay-inclusive haptic communication dynamics model and derive both delay-independent and delay-dependent frequency-domain stability criteria, validated through theoretical analysis, numerical simulations, and real-robot experiments. Contribution/Results: We quantitatively characterize the nonlinear dependence of the maximum tolerable delay on controller parameters, robotic stiffness, and actuator performance, establishing critical delay thresholds across operational conditions. This work presents the first quantitative analysis of the coupling between stability and motion performance—specifically, how stability margins directly constrain tracking accuracy and transparency. The results provide a general theoretical foundation and experimentally verifiable design guidelines for robust control and delay-compensation strategies in tele-collaborative haptic systems.

Wearable multi-limb systems impose external torques on the human body during operation, compelling additional motor units to engage in postural regulation, thereby compressing the muscle null space and degrading human–robot collaboration efficiency and comfort. To address this, we propose a torque-suppression method based on coordinated motion planning: angular acceleration constraints and soft position-error limits are incorporated at the motion-planning layer, enabling online trajectory optimization to dynamically avoid high-torque configurations; concurrently, a simplified human–robot coupled dynamic model is employed to minimize peak joint torques while ensuring dynamic feasibility. Simulation results demonstrate that the method reduces disturbance torques on the torso and supporting limbs by 32.7% on average, expands the muscle null space by approximately 28%, and enhances operational safety and interaction naturalness. The core innovation lies in explicitly embedding torque suppression into the motion-planning layer, enabling proactive mitigation of human–robot dynamic interference.

This study addresses the challenges of high cognitive load on single operators and substantial coordination overhead in multi-operator setups for multi-arm teleoperation tasks. The authors propose a human–robot collaborative teleoperation framework wherein a human directly controls two master arms, while two auxiliary arms are autonomously managed by a training-free multimodal large language model (MLLM) agent to execute subtasks, with real-time intervention enabled via voice commands. This approach pioneers the integration of MLLM agents into multi-arm teleoperation for data collection, achieving decoupled control spaces and natural human–robot interaction. The system maintains high operational efficiency while significantly enhancing scalability. Experimental results demonstrate that the proposed method achieves data collection success rates and efficiency comparable to those of expert two-human teams, and the collected data effectively supports downstream training of multi-arm collaborative policies.