A GPU-Based Evolution Algorithm for Motion Planning of a Redundant Robot

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Kinematic redundancies of robotic manipulator provide many benefits for motion planning, such as collision avoidance, improved manipulability, flexibility, and singularity avoidance. Based on the task priority, the motion planning of a redundant robot involves solving the associated optimization problem according to the desired criterion. Several kinematic techniques for redundant manipulators have been proposed. However, the computing time of solving the optimization problem with multiple optimization objectives is too large to implement the motion planning in real-time. Therefore, more efficient optimization algorithms are needed for motion planning of redundant robots. This paper presents a new technique to solve the inverse kinematics of redundant manipulators using a GPUbased evolution algorithm. This algorithm combines the optimal perturbation method with a multi-objective function to solve the obstacle avoidance and trajectory tracking at the same time. Simulations and implementations for a selfdesign manipulator with eight joints considering the optimization with multiobjectives with obstacle avoidance and trajectory tracking are developed. The results reveal that the proposed evolution algorithm based on compute unified device architecture (CUDA) has the much higher performance than the optimal perturbation method. Simulation results also show that the proposed EAMP is 398 times faster than the perturbation method for this case study. Experimental results using NI Compact RIO® validate the proposed method feasible for realtime motion planning of the redundant robot.

This publication has 31 references indexed in Scilit: