Skip to content

Conversation

@Sudhendra
Copy link

In Light Thief (the env), the agent must navigate a dark arena to collect loot. The core mechanic features a "visibility twist": loot is only revealed when illuminated by moving searchlights, but the agent can only collect it when they are in total darkness. This requires the agent to memorize loot locations and time their movements to avoid detection while staying close enough to track their targets.

Planned improvements:

  1. Box2D Integration: Transition from simple circle-based physics to full Box2D for complex collisions and obstacles.
  2. WebAssembly Support: Finalize the emscripten build pipeline for browser-based demos.
  3. Procedural Generation: Add randomized level layouts, light patterns, and loot placements to improve agent generalization.
  4. Multi-Agent Support: Extend the environment for competitive or cooperative multi-agent scenarios.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant