RoverEdge

RoverEdge is an evidence-driven robotics experimentation environment for understanding how AI inference latency affects autonomous inspection missions. Users can run and replay simulated rover missions, model edge-device compute delays, and compare runtime configurations through controlled experiments. When a mission fails, a Meta model served through Baseten acts as a bounded mission engineer: it examines the recorded trace, proposes one supported configuration change, and explains whether the follow-up experiment helped. RoverEdge executes every experiment locally and validates the model’s conclusion against recorded evidence. This prevents unsupported claims of improvement and clearly separates AI-generated reasoning from deterministic simulation results. The project includes an interactive 3D replay, modeled queue behavior, observation-freshness evidence, reproducible experiment bundles, and portable hosting support through Modal. Its goal is to help robotics engineers move from “the rover failed” to a precise, testable, and evidence-backed explanation.