Obstacle avoidance algorithm using gradient based Swarm techniques

摘要

In this paper, a hybrid approach to obstacle avoidance, based on Particle Swarm Optimisation is proposed. This method provides significantly faster convergence, compared to classical approaches using potential fields and gradient descent. The potential functions being used are presented, along with the results, one would obtain by employing gradient descent, for comparison. The results obtained by using hybrid-algorithm, clearly show the significant reduction in number of iterations taken for convergence, in comparison to the exponential time, typically taken by gradient descent. The penultimate section explains the approach taken to adapt the algorithm being proposed, for applications with GPS coordinates. Experimental results for the same are also presented herewith.