Proposal and Evaluation of a Tabu List Based DQN for AAV Mobility

Abstract

The Deep Q Network (DQN) is one of the methods of the deep reinforcement learning algorithm, which is a deep neural network structure used to estimate Q-values in Q-learning methods. The authors have previously designed and implemented a DQN-based mobility control methods for Autonomous Aerial Vehicle (AAV). In this paper, we propose and evaluate a DQN based on tabu list strategy for AAV mobility control. For evaluation, we simulate were conducted for the mobility control of AAV in a staircase environment using normal DQN and tabu list based DQN. The simulation results showed that a tabu list based DQN was a better solution than the normal DQN.