comeanotherday4/mutil_agent_path_planning

This research project compares reinforcement-learning methods for resource-constrained desert route planning across three scenarios. Users run a scenario script to simulate travel, resource purchases and mining under weather and deadline constraints, then inspect learned or optimized routes and outcomes. Scenario 1 compares dynamic programming with Q-learning for known weather; scenario 2 evaluates Q-learning under four initial resource mixes; scenario 3 compares fixed-policy-plus-PPO, IPPO and MAPPO in a three-agent setting. The source excerpts are partial, so links between unsampled implementations are omitted; the DP capability is documented in the README.