Online Optimization of Movement Cost for Robotic Applications of PSO

Abstract

Particle Swarm Optimization is an optimization algorithm that can be used as a control mechanism in robotic tasks, especially robotic search. Existing algorithms are tuned to use as little evaluations of the objective function as possible. Measuring the objective with a sensor usually has a low cost in terms of time and energy compared to moving the robot. We propose a new algorithm to optimize the particle movement in SMPSO that samples the same points in the environment with less movement cost. Our experiments show that the average movement cost can be reduced by $$50\%$$ or more in all test problems we used. The huge gain shows that there is a big potential in adapting swarm intelligence algorithms to robotic applications by optimizing them to better serve the constraints of the application.