A Hybrid Reinforcement Learning and PSO Approach for Route Discovery of Flying Robots

Abstract

Route discovery for flying robots is one of the major concerns while developing an autonomous aerial vehicle. Once a path planning algorithm is built and the flying robot reaches the destination point from the target point successfully, it is again important for the flying robot to come back to its original position and that is done through route discovery algorithms. Reinforcement learning is one of the popular machine learning methods in which the flying robot has to interact with the environment and learn by exploring the possibilities and maximum reward point method, without the requirement of a large amount of prior training data. Particle swarm optimization is an artificial intelligence inspired algorithm which finds optimal solution in a multi-dimensional space. This chapter has discussed a random exploration reinforcement learning approach combined with PSO algorithm that has been used to discover the optimum path for a flying robot to return from the destination point to the target point after it had traversed its best path from an already defined swarm intelligence technique. PSO+Reinforcement Learning (RL-PSO) is an optimization technique that combines the global search capability of PSO with the exploitation and exploration strategy of RL. Here higher reward points were assigned to the already defined best path obtained from the path planning technique, so that while returning from the destination point it will try to find the route with the highest reward point. With several iterations, it will optimize and find the best route for backpropagation. The algorithm is built using a python environment and the convergence result with the number of iterations has been validated.