Non-Myopic Neuro-Symbolic AI for Smart Infrastructure

Invited Talk at Nanyang Technological University (NTU), Singapore — Abhishek Dubey

A comprehensive overview of the SCOPE Lab's neuro-symbolic approach to decision-making in societal-scale cyber-physical systems. The talk covers how combining neural perception (demand forecasting, anomaly detection, state estimation) with symbolic reasoning (constraint satisfaction, formal planning, compositional logic) and non-myopic planning enables AI systems that are simultaneously fast, safe, and far-sighted. Includes deployed system case studies from Chattanooga and Nashville — vehicle-to-building energy coordination, microtransit scheduling, school bus operations, and transit disruption management — as well as recent methodological advances in adaptive Monte Carlo tree search (ADA-MCTS), particle filtering for POMDPs (ESCORT, AIR-POMDP), learned temporal abstraction (L-MAP, ICLR 2025 Spotlight), and the NS-Gym benchmarking framework for non-stationary environments.

neuro-symbolic AI cyber-physical systems transit optimization energy systems MCTS POMDPs
March 2026