code for `Autonomous navigation of UAV in multi-obstacle environments based on a Deep Reinforcement Learning approach' Google Scholar
Video: YouTube
- Unity 2019.3.2f1
- Unity Hub 2.3.2
- ml-agents 0.15.1
- Anaconda
- numpy, torch, gym_unity, etc.
- UAV Project: `Drone Controller Full' (now unavailable, and replacement could be `Drone Controller Pro') in Unity Asset Store.
- TD3: Policy-Gradient-Methods
open rl/TD3_original/TD3_original.ipynb in jupyternotebook and run it.
Notice:
Cell 3 is only for examining if the environment works well
Cell 5 is only for training
Cell 9 is only for testing
If you wish to modify the environment settings, including DRL settings (actions, observations, rewards, terminal conditions), collision detection, laser ranging, etc, you can find them in Assets/RocketAgent.cs.
Or if you're interested in creating a new environment, the guide `Learning-Environment-Create-New' could be helpful.