Obstacle Avoidance for Aerial Vehicles in Autonomous Navigation

Obstacle avoidance is the back bone of autonomous navigation as it enables vehicles to reach desired location avoiding hurdles in the path. It is one of the ongoing challenging researches in the arena of cyber physical systems. In this article, comparison of various obstacle avoidance algorithms such as Artificial Potential Field (APF) approach, Vector Field Histogram (VFH) approach, Bubble Band approach, Mounted Sonar approach, Dist-Bug approach, bug-1 and bug-2 approach and Tangent Bug approach has been addressed. The main objective is obstacle avoidance, where obstacle can be in the form of radars or any other type of equipment. A novel obstacle avoidance procedure for low-altitude flying vehicles (Static Obstacle Avoidance) and high-altitude flying vehicle (Dynamic Obstacle Avoidance) has been proposed using A-Star and Deep Q-Network Reinforcement Learning techniques, respectively. Test bed has been created considering object model, sensor model, obstacle environment and controller. Implementation of the same has been done through visualization using Pygame for A-Star approach and Director for Deep Q-Network Reinforcement Learning.