Unity Indoor Scene + ML Agents
Budget: ₹1,500 – ₹12,500 INR
I have already licensed and downloaded a pack of Unity-ready models from the Asset Store; what I now need is a clean, functional indoor scene built from those assets and wired for research with Unity ML-Agents.
The job breaks down into two parts:
1. Scene assembly
• Arrange the supplied meshes into a believable indoor layout (corridors, racks, etc.).
• Optimise lighting, textures, colliders and a baked NavMesh so performance is solid on mid-range GPUs.
2. ML-Agents integration
• Drop in the latest ML-Agents package, set up the behaviour parameters and create stubs so I can plug in my own training code.
• Add interactive objects—specifically autonomous moving bots—so agents have dynamic elements to sense and learn from. Basic path-finding and event hooks (e.g., OnBotSeen, OnCollision) are enough; I’ll handle the learning logic myself.
Acceptance criteria
• Opening the project in Unity 2022 LTS runs without errors.
• Pressing Play spawns the bots, allows navigation through the indoor space and streams observations/actions through the ML-Agents framework.
• Scene stays above 60 FPS in the editor on a GTX 1660 Ti or better.
Deliverables
• Complete Unity project folder (URP preferred but not mandatory).
• Clear README explaining where to place additional datasets or scripts I may add later.
If you have prior examples of indoor ML-Agents setups or moving bot AI, please include a short clip or repo link so I can gauge fit quickly.
The job breaks down into two parts:
1. Scene assembly
• Arrange the supplied meshes into a believable indoor layout (corridors, racks, etc.).
• Optimise lighting, textures, colliders and a baked NavMesh so performance is solid on mid-range GPUs.
2. ML-Agents integration
• Drop in the latest ML-Agents package, set up the behaviour parameters and create stubs so I can plug in my own training code.
• Add interactive objects—specifically autonomous moving bots—so agents have dynamic elements to sense and learn from. Basic path-finding and event hooks (e.g., OnBotSeen, OnCollision) are enough; I’ll handle the learning logic myself.
Acceptance criteria
• Opening the project in Unity 2022 LTS runs without errors.
• Pressing Play spawns the bots, allows navigation through the indoor space and streams observations/actions through the ML-Agents framework.
• Scene stays above 60 FPS in the editor on a GTX 1660 Ti or better.
Deliverables
• Complete Unity project folder (URP preferred but not mandatory).
• Clear README explaining where to place additional datasets or scripts I may add later.
If you have prior examples of indoor ML-Agents setups or moving bot AI, please include a short clip or repo link so I can gauge fit quickly.