Neuroadaptive Fixed-Time Synchronous Control With Composite Learning Policy for Robotic Multifingers
提出一种神经网络复合学习策略，实现拟人多指手在未知动力学和干扰下的固定时间同步控制，通过图论描述手指间连接，利用历史数据和回归矩阵构造预测误差，在较弱的区间激励条件下实现精确参数估计。
Dexterous manipulation of anthropomorphic multifinger robotic hands (MFRHs) is crucial for performing diverse and intricate tasks, where collaboration among the fingers is essential. This article presents a novel neural network-based composite learning strategy tailored for the synchronous control of multiple fingers in anthropomorphic MFRHs subjected to unknown dynamics and disturbances. By leveraging graph theory, the interconnections among fingers are delineated and integrated into the dynamic equations. The modified nonsingular terminal sliding mode (TSM) technique is employed to achieve fixed-time convergence of error variables without triggering singularity. Within the framework of composite learning, a novel computable prediction error is formulated by harnessing online historical data alongside the regression matrix. The combination of prediction errors and the regression matrix is utilized for parameter estimation, which, under a milder interval excitation (IE) condition, facilitates accurate parameter estimation without the requirement for the stringent persistent excitation (PE) condition. The feasibility and effectiveness of the proposed technique are demonstrated through simulation experiments.