Investigation into the decoding of genetic-based robot motion considering sequential and parallel formulations

The use of genetic algorithms is investigated for the solution of a computationally intensive problem, the motion planning of multi-joint arms. Previous reports have indicated continued difficulties with real-time implementation of the GA scheme, especially for industrial arms with 6 DOF. In this paper, parallel genetic algorithms are investigated as a potential solution, and more consideration is given to the decoding of the chromosomes. The reported scheme makes it possible to reduce the number of searched generation significantly, thus making it more feasible to implement in real-time. Simulation results for the motion planning of a six DOF PUMA arm is included, demonstrating the applicability of the scheme.