Motion planning for redundant robotic manipulators using a novel multi-group particle swarm optimization

Metaheuristic optimization algorithms are widely used in motion planning of redundant robotic manipulators. Existing methods may converge to a local minimum. In this paper, a new multi-group particle swarm optimization algorithm (PSOEL) is proposed to solve the motion planning of manipulators. PSOEL consists of one elite group and several child groups. The population is initialized with a pre-selection mechanism in which the members of the elite group are initialized with the best-performing particles of the child groups. In the process of iteration, the elite group and the child groups evolve separately. When the elite group falls into a local optimum or is inferior to child groups for a certain time, an interaction mechanism is triggered. In the interaction mechanism, some of the best particles selected from the child groups will replace the bad particles of the elite group. With these mechanism of pre-selection and interaction, PSOEL can jump out of the local optimum and reach the global optimum or global suboptimum. Simulation results show that the proposed algorithm PSOEL is superior to the compared algorithms and converges toward the optimum.