This project demonstrates an Autonomous Player capable of navigating a dynamic environment within Unity, leveraging advanced Reinforcement Learning (RL) techniques. Using Python, Gymnasium, and Unity's C# scripting, the agent learns to efficiently move toward randomized targets while avoiding obstacles.
-
Custom Gym Environment: Built using Gymnasium and integrated seamlessly with Unity.
-
PPO Algorithm: Efficient learning via Stable-Baselines3's Proximal Policy Optimization.
-
RPC Communication: Robust JSON-RPC integration between Unity (C#) and Python scripts.
-
Dynamic Navigation: Autonomous behavior in response to randomized goals and obstacle placement.
-
Unity Engine (C# scripting)
-
Python with Gymnasium
-
Stable-Baselines3 (PPO algorithm)
-
peaceful_pie (Unity-Python RPC communication), Thanks to Hugo RL
git clone <repo-link>
pip install gymnasium stable-baselines3 peaceful_pie
Open Unity project, import necessary scripts, and ensure the correct RPC ports are configured.
python train_rl.py --port 9000
python test_rl.py --model PPO_Test_Model_1 --episodes 50
Unity runs an RPC server, exposing methods that can be called remotely. Methods in Unity are marked with [JsonRpcMethod] attributes in your C# scripts. // Unity side (RPC server)
[JsonRpcMethod] RlResult Step(string action) { return new RlResult(reward, finished, truncated, GetObservation()); }
Python scripts connect to Unity's RPC server using the UnityComms class provided by Peaceful Pie. Methods defined in Unity can be directly invoked from Python, allowing Python to issue commands like moving an object or obtaining game state information. Python side
unity_comms = UnityComms(port=9000)
Calling Unity's Reset method
initial_obs = unity_comms.Reset(ResultClass=MyVector3)
Taking a Step in the environment by calling Step method on Unity side
result = unity_comms.Step(action='north', ResultClass=RlResult)
// Sooon Picture will be here.
The trained agent successfully navigates dynamically generated environments, effectively demonstrating RL's capability to adapt and improve navigation strategies autonomously.
Contributions, issues, and feature requests are welcome! Feel free to check the issues page or open a pull request.
None - Free to Use & Fork
Made with ❤️ by Mansoor