Brain-Inspired Strategy for the Motion Planning of Hyper-Redundant Manipulators

Liangliang Zhao,Jingdong Zhao,Hong Liu
DOI: https://doi.org/10.1109/cyber.2016.7574834
2016-01-01
Abstract:The main challenge of motion planning for a hyper-redundant manipulator is to implement a modular structure ensure real time and high performance of the control system. In this research, we present a strategy to deal with the motion planning problem of a hyper-redundant manipulator, include uncertain time delay to the control system and obstacle avoidance. Similarly to the principles of motor control in human brain, we extract primitive motions from a batch of motion data of a hyperredundant manipulator, and reprogram the complex motions by the sequence of combinations of primitive motions. Based on the neural network algorithm, we present the simulation results of training experiment and testing experiment. All the simulations have confirmed that the proposed control strategy provides remarkable efficiency in motion planning of hyper-redundant manipulators.
What problem does this paper attempt to address?