ARES-MAS

ARES-MAS is a simulated, grid-based orchestrator designed to manage collaborative planetary exploration through a hybrid symbolic AI architecture. The project addresses the complex challenges of autonomous multi-agent coordination, specifically focusing on dynamic hazard avoidance, spatial pathfinding, and safety-critical resource management. To achieve this, the system enforces a strict architectural separation of concerns. The physical layer, implemented in Kotlin, provides a discretetime 2D spatial simulator that manages environmental state, battery consumption, and mutually exclusive resource locks. The deliberative logical layer encapsulates the system’s intelligence utilizing BDI (Belief-Desire-Intention) agents (Scouts and Harvesters) programmed in Jason and AgentSpeak(L). These autonomous agents communicate via the FIPA Agent Communication Language (ACL). By leveraging the Contract Net Protocol (CNP) for task negotiation, the agents evaluate real-time spatial heuristics (Manhattan distance) to effectively orchestrate task assignment. Crucially, a logic-based safety layer utilizing tuProlog acts as a runtime guardrail, preventing fatal collisions and ensuring fail-safe operations when agents make overoptimistic economic decisions.