CMU Develops AI to Predict Airport Collision Risks Using PSC's Bridges-2
Researchers from CMU's AirLab and the PSC have developed World2Rules, an AI designed to predict and explain airport collision risks. Utilizing Bridges-2, the AI analyzes airport data to enhance safety in aviation operations, representing a significant advancement in collision risk management.
Theia Market Signal Identification - AI AssistedThe World2Rules AI, developed by CMU's AirLab, aims to predict and explain collision risks in airport operations using PSC's Bridges-2 supercomputer. The AI integrates data from the Amelia-42 dataset, which contains approximately 10 terabytes of surface movement data from 42 U.S. airports, and combines neural and symbolic AI methods for enhanced interpretability.
Initial testing shows a 23.6% improvement over purely neural approaches in learning accurate safety rules. Future enhancements will focus on incorporating evolving data to better address uncertainties in vehicle movements. The AI's framework could also be applicable in various safety-critical domains beyond aviation.
Comments