How soofiyevv/parkingagent works
ParkingAgent is a Unity ML-Agents reinforcement-learning project: cars observe a randomized parking scene, learn to park through PPO training, and can run the exported policy in Unity. The documented workflow connects the scene, agent, trainer, configurations, and model; source excerpts were not available, so implementation details beyond the README are omitted.
Main entry points: SampleScene.unity Parking scene · CarAgent.cs Car agent · CarParking.onnx Trained policy · car_parking.yaml Baseline PPO
Components 10
Parking simulation
- Parking scene
Assets/Scenes/SampleScene.unity - Car agent
Assets/Scripts/CarAgent.cs - Ray sensors
Assets/Scenes/SampleScene.unity
PPO training
- Baseline PPO
config/car_parking.yaml - Small PPO
config/car_parking_small.yaml - High-rate PPO
config/car_parking_lr.yaml - ML-Agents trainer
Policy inference
- Trained policy
Assets/Models/CarParking.onnx
Other components
- Project user
- TensorBoard
Connections 15
- Parking scene → Car agent: hosts agents
- Car agent → Ray sensors: uses observations
- Ray sensors → Car agent: provides rays
- Car agent → Parking scene: controls car
- Project user → ML-Agents trainer: starts training
- ML-Agents trainer → Baseline PPO: reads config
- ML-Agents trainer → Small PPO: reads config
- ML-Agents trainer → High-rate PPO: reads config
- ML-Agents trainer → Parking scene: runs environment
- Parking scene → ML-Agents trainer: sends experience
- ML-Agents trainer → Trained policy: exports policy
- Project user → Parking scene: runs inference
- Parking scene → Trained policy: loads policy
- ML-Agents trainer → TensorBoard: logs metrics
- Project user → TensorBoard: monitors training