Nonlinear State Estimation and Online Neighbor Selection for Multimanipulator Systems

Abstract

In this article, we focus on the position synchronization of nonlinear multimanipulator systems. For robot manipulators that are only equipped with joint position measurement devices with measurement noises, a continuous-discrete adaptive unscented Kalman filter (CD-AUKF) is implemented to acquire smoother manipulator position states, and, meanwhile, estimate high-order states (e.g., velocity and acceleration). However, in closed-loop control of networked multimanipulator systems, using estimated states may result in a drastic increase in tracking errors. This shortcoming is addressed by an energy index-based neighbor selection policy (NSP). To maintain a higher tracking performance, the NSP allows each agent to actively select well-performing neighbors to interact with, while the poor-quality neighbor data is discarded. Finally, to regulate the multimanipulator system in the presence of parametric uncertainties, friction, disturbances, time-varying network delays, and packet loss, an adaptive nonsingular terminal sliding-mode (ANTSM) controller is designed. A group of Phantom Omni robotic devices were used to carry out numerical simulations and experimental studies that demonstrate the effectiveness of the proposed ANTSM control method, CD-AUKF estimation, and active neighbor-selection policy.