UAV Target Tracking Method Based on Deep Reinforcement Learning
- Resource Type
- Conference
- Authors
- Zhang, Haohui; He, Pingkuan; Zhang, Ming; Chen, Daqing; Neretin, Evgeny; Li, Bo
- Source
- 2022 International Conference on Cyber-Physical Social Intelligence (ICCSI) Cyber-Physical Social Intelligence (ICCSI), 2022 International Conference on. :274-277 Nov, 2022
- Subject
- Communication, Networking and Broadcast Technologies
Computing and Processing
General Topics for Engineers
Power, Energy and Industry Applications
Robotics and Control Systems
Signal Processing and Analysis
Transportation
Deep learning
Visualization
Target tracking
Navigation
Atmospheric modeling
Neural networks
Reinforcement learning
GRU
TD3
UAV target tracking
visual navigation
- Language
This study proposes a UAV target tracking method using reinforcement learning algorithm combined with Gate Recurrent Unit (GRU) to promote UAV target tracking and visual navigation in complex environment. Firstly, an algorithm Twins Delayed Deep Deterministic policy gradient algorithm (TD3) using deep reinforcement learning and the GRU gated loop unit are introduced. The unit is then added to the neural network to process continuous time data, and the algorithm TD3 is adopted to train the model so that it can drive the UAV to make autonomous flight decisions and accomplish target tracking. The proposed method is verified on the AirSim simulation platform.