Ayushi Chadha

Factory Recovery Lab is a simulation-grounded autonomous recovery system for manufacturing environments. It explores a central question in agentic decision-making: when should an AI supervisor preserve an existing plan, investigate alternatives or intervene? The project simulates a factory with interconnected production operations, machine failures, repair uncertainty and delivery deadlines. A Meta-powered supervisory agent observes disruptions, requests recovery experiments and decides whether to persist with the current schedule or select a validated alternative. The system combines Muse Spark with CP-SAT scheduling, SimPy simulation and independent constraint validation. The agent can compare recovery strategies, investigate uncertainty through stress tests and examine their predicted consequences before making a decision. Muse Spark vision and Muse Voice Transcribe also enable reviewed maintenance reports to inform new simulation scenarios. Rather than simply automating factory scheduling, the project provides an experimental environment for studying supervisory control, intervention costs and decision-making under uncertainty. Its broader goal is to understand when autonomous agents should act, when they should gather more evidence and when maintaining an existing strategy is the better decision.