Autonomous Landing Logic and Performance Evaluation

Develop the autonomous decision-making logic that uses the real-time hazard detections from your `TorchServe` endpoint to identify safe landing zones and, if necessary, suggest avoidance maneuvers or alternative landing trajectories. Incorporate a mechanism to interpret dynamic mission constraints provided as natural language (e.g., via `Phi-3`'s capabilities) such as 'prioritize landing areas with sun exposure' or 'avoid slopes exceeding X degrees.' Describe how you would simulate and evaluate the end-to-end system's performance in a dynamic lunar descent scenario, including metrics like safe landing success rate and decision latency.