How thredgroup/multi-agent-robots works
This Unity project demonstrates multi-agent reinforcement learning for intelligent manufacturing with autonomous mobile robots. The principal workflow is a Unity manufacturing simulation involving robot, machine, and operator agents, with PPO configuration and trained-policy artifacts supporting learning and evaluation. The README identifies these scripts and artifacts, but no source excerpts were supplied, so the graph avoids inferring runtime calls or data flow between them.
Main entry points: MobileRobotAgent.cs Mobile Robot · Machines.cs Machine Agent · OperatorAgent1.cs Operator Agent 1 · OperatorAgent2.cs Operator Agent 2
Components 10
Manufacturing Simulation
- CityCombat Scene
Agent Roles
- Mobile Robot
Assets/MobileRobotAgent.cs - Machine Agent
Assets/Machines.cs - Operator Agent 1
Assets/OperatorAgent1.cs - Operator Agent 2
Assets/OperatorAgent2.cs
Learning Artifacts
- PPO Settings
config/ppo/multi-agent-robots-config.yaml - Trained Network
Assets - Policy Recordings
Recordings
Other components
- Researcher
- ML-Agents Toolkit
Connections 8
- Researcher → CityCombat Scene: runs simulation
- Researcher → PPO Settings: sets PPO
- ML-Agents Toolkit → Mobile Robot: supports training
- ML-Agents Toolkit → Machine Agent: supports training
- ML-Agents Toolkit → Operator Agent 1: supports training
- ML-Agents Toolkit → Operator Agent 2: supports training
- Researcher → Trained Network: uses policy
- Researcher → Policy Recordings: reviews results