UME: A Unified Meta-Generalization Framework for Cross-Domain ETA
Researchers have developed UME, a Unified Meta-generalization framework designed to improve Estimated Time of Arrival (ETA) predictions in logistics. This framework addresses challenges like generalizing to new domains, handling missing features in emerging markets, and enabling knowledge transfer across different regions. UME integrates a dual-branch architecture with a meta-learning mechanism to dynamically adjust predictions based on domain knowledge and instance context, and has been successfully deployed on the Meituan-keeta delivery platform. AI
IMPACT Enhances logistics efficiency and user satisfaction through improved ETA predictions in diverse operational environments.