An approach for UAV indoor obstacle avoidance based on AI technique with ensemble of ResNet8 and Res-DQN

Unmanned Aerial Vehicles (UAVs) have many important applications in both civil and military areas. One of the most popular type of UAVs is quad-copter which uses four propellers to carry out its flight process. In this paper, a new control model which helps quad-copter to automatically find path and avoid obstacles indoor is introduced. The challenge of this model for quad-copter is the complex indoor environments with obstacles. Base on Deep Reinforcement Learning and Deep Learning platform, state of the art algorithms in Artificial Intelligence (AI), a new Ensemble model is proposed. The proposed model uses two algorithms to control quad-copter. One is quad-copter path finding algorithm (Deep Learning - ResNet8) and the other algorithm (Deep Reinforcement Learning - Res-DQN) dealing with obstacle avoidance. The output of both two algorithms are combined to change the direction of quad-copter adaptively with indoor environments. The simulation results have been assessed to verify the numerous performance of proposed control model.