Obstacle avoidance with DQN Algorithm (Double DQN, Prioritized Experience Replay, Dueling Network Architectures, N-step Return, Soft TargetNet, Parallel computing...
In the given environment, the agent/white dot need to avoid collision with obstacles/black dots.
A3C algorithm from: https://github.com/jaromiru/AI-blog
python train_dqn_v*.py
I uploaded the saved weights in folder weights_saved.
