导纳
控制理论(社会学)
计算机科学
控制工程
控制(管理)
自适应控制
动态规划
控制系统
振动控制
执行机构
气动流量控制
气动人工肌肉
振动
工程类
运动控制
人工神经网络
气动执行机构
机器人
加速度
最优控制
车辆动力学
作者
Xiaohong Cui,Jielei Shao,Zimeng Wan,Guang‐Hong Yang,Binrui Wang,Quanmin Zhu
标识
DOI:10.1109/tie.2026.3718192
摘要
Pneumatic artificial muscles (PAMs) are promising actuators for robotic gripping due to their lightweight structure and intrinsic compliance. However, achieving accurate position and force control is challenging because of their strong nonlinearities, hysteresis, and time-varying dynamics, particularly in contact-rich manipulation tasks. This article proposes a hierarchical force/position control framework for a PAM-driven gripper by integrating adaptive dynamic programming (ADP) with admittance control. The outer-loop admittance controller generates reference corrections based on the force-tracking error, while the inner-loop ADP controller tracks the updated reference despite model uncertainty. Using the nominal dynamics, the associated optimal control problem is formulated within the Hamilton–Jacobi–Bellman framework, and the control policy is approximated online using a neural network (NN). Lyapunov analysis shows that all closed-loop signals are uniformly ultimately bounded. Simulation results indicate that, compared to a proportional-integral-derivative (PID)-based admittance controller, the proposed method achieves reduced overshoot and improved steady-state force-tracking accuracy. Experiments on a PAM-driven gripper further validate the feasibility and effectiveness of the proposed framework for stable force regulation and satisfactory tracking in grasping tasks.
科研通智能强力驱动
Strongly Powered by AbleSci AI