JustMe

REROUTE is an uncertainty-aware autonomous AI agent designed to help intelligent systems adapt when unexpected situations occur. Traditional autonomous systems often depend on predefined plans and struggle when actions fail, obstacles appear or environmental information is incomplete. REROUTE addresses this challenge through an intelligent perception, reasoning, decision-making and recovery pipeline. The system continuously evaluates its environment and execution state, detects failures, diagnoses their causes and evaluates possible recovery actions. It integrates Meta Muse Spark through the Meta Model API to provide AI-powered reasoning using structured context and visual evidence. REROUTE then evaluates the model's recommendations alongside feasibility, risk and predicted outcomes before selecting an action. Unlike systems that simply repeat failed actions, REROUTE can replan, gather additional information, wait or request human assistance. It also verifies whether recovery actions succeed and whether the original objective has been achieved. The project is implemented in Python and demonstrated using a PyBullet robotics simulation with A* path planning, obstacle navigation and autonomous failure recovery. Our goal is to advance autonomous systems from simply following instructions to intelligently adapting when the unexpected happens.