A Modified Q-learning Algorithm Facing for Path Planning and Obstacle Avoidance Method for Unmanned Aerial Vehicle

Nowadays, Unmanned Aerial Vehicles (UAVs) have been widely used in the area of aerial photography, information collection during emergencies, and goods transportation. Most path planning algorithms require a map of the operation area such that an obstacle-free path can be solved. Path planning and obstacle avoidance become very challenging. In this paper, a variable learning rate Q-learning algorithm for path planning and obstacle avoidance problem is proposed. This algorithm can avoid getting stuck in exploration-exploitation dilemmas and local deadlock states, which is often encountered with the classic Q-learning algorithm. This is accomplished by incorporating state-action variables, employing a variable step size update method, implementing an epsilon-greedy strategy, considering a distance weight factor, and introducing randomness to expedite convergence. Finally, the simulation is conducted to show the superior performance of the proposed algorithm in its convergence speed and rewards.