MIT engineers have developed an approach to quickly identify potential failures in autonomous systems and suggest repairs to avoid system breakdowns. The approach has been tested on various simulated systems, including power grids, aircraft collision avoidance systems, rescue drones, and robotic manipulators. By taking a sensitive approach and generating random changes within the system, the algorithm can identify a wider range of failures and potential fixes. The researchers believe this approach can be applied to any autonomous system with an accurate simulation.
