Q learning algorithm based UAV path learning and obstacle avoidence approach

• Published in: Chinese Control Conference pp. 3397 - 3402
• Main Authors: Zhao Yijing, Zheng Zheng, Zhang Xiaoyi, Liu Yang
• Format: Conference Proceeding
• Language: English
• Published: Technical Committee on Control Theory, CAA 01.07.2017
• Subjects: Artificial intelligence; Collision avoidance; History; neural network; Neural networks; Path planning; q learning; Real-time systems; trap-escape strategy; UAV obstacle avoidance; Unmanned aerial vehicles
• DOI: 10.23919/ChiCC.2017.8027884

As Unmanned Aerial Vehicle (UAV) having been applied in more complex and adverse environments, the requirements of automatic techniques for obstacle avoidance are becoming more and more important. Reinforcement learning (RL) is a well-known technique in the domain of Machine Learning (ML), which interacts with the environment and learning the knowledge without the requirement of massive priori training samples. Thus it is attractive to implement the idea of RL to support UAV tasks in unknown environments. This paper adopts an Adaptive and Random Exploration approach (ARE) to accomplish both the tasks of UAV navigation and obstacle avoidance. Search mechanisms will be conducted to guide the UAV escape to a proper path. Simulations on different scenarios show that our approach can effectively guide UAVs to reach their targets in quite rational paths.