Fall prediction, control, and recovery of quadruped robots
Hao Sun,Junjie Yang,Yinghao Jia,Chong Zhang,Xudong Yu,Changhong Wang
DOI: https://doi.org/10.1016/j.isatra.2024.05.039
IF: 7.3
2024-05-26
ISA Transactions
Abstract:Highlights • Capturability-based online fall prediction algorithms are proposed for planar single-contact and 3D multi-contact locomotion. The incorporation of centroidal angular momentum and multiple foot placements into the system dynamics enhances the descriptive nature of capturability for quadrupedal locomotion. Compared with previous methods, our proposed fall prediction approach exhibits significantly improved reliability and flexibility. • By incorporating modeling uncertainty into the system dynamics, a novel contact-implicit trajectory optimization method is proposed to avoid falling or reduce falling damage. This method effectively addresses the non-smoothness of contact dynamics while enhancing the robustness of resulting trajectories. Unlike existing fall control methods that use simplified models and pre-specified contact sequences, the proposed controller not only generates state and input trajectories but also provides contact mode sequences for safe fall control. • A model-free deep reinforcement learning-based approach is presented to complete the fall recovery task. Compared with our previous work, our learning algorithm overcomes the limitation of a single-stage reward, which hinders the learning of roll-over and standup behaviors for fall recovery in 12-DOF quadruped robots. Compared with existing algorithms that may suffer from insufficient or redundant explorations, the proposed learning framework automatically discover initial states to improve data efficiency and robustness. When legged robots perform complex tasks in unstructured environments, falls are inevitable due to unknown external disturbances. However, current research mainly focuses on the locomotion control of legged robots without falling. This paper proposes a comprehensive decision-making and control framework to address the falling over of quadruped robots. First, a capturability-based fall prediction algorithm is derived for planar single-contact and 3D multi-contact locomotion with a predefined gait sequence. For safe fall control, a novel contact-implicit trajectory optimization method is proposed to generate both state and input trajectories and contact mode sequences. Specifically, incorporating uncertainty into the system and terrain models enables mitigating the non-smoothness of contact dynamics while improving the robustness of the resulting trajectories. Furthermore, a model-free deep reinforcement learning-based approach is presented to achieve fall recovery after the robot completes a fall. Experimental results demonstrate that the proposed fall prediction algorithm accurately predicts robot falls with up to 95% accuracy approximately 395ms in advance. Compared to classical locomotion controllers, which often struggle to maintain balance under significant pushes or terrain perturbations, the presented framework can autonomously switch to the fall controller approximately 0.06s after the perturbation, effectively preventing falls or achieving recovery with a threefold reduction in touchdown impact velocity. These findings highlight the effectiveness of the proposed framework in enhancing the stability and safety of legged robots in unstructured environments.
automation & control systems,instruments & instrumentation,engineering, multidisciplinary