Metropolis-Scale Resilient and Trustworthy Traffic Flow Inference Using Multi-Source Data
Researchers have developed a new probabilistic framework called the Task-Aware Attentive Neural Process (TA-ANP) for inferring traffic flow across large urban areas using sparse data. This method fuses floating car data with fixed sensor measurements, adapting to changing sensor configurations without retraining. TA-ANP addresses challenges like underdetermined problems and conflicting inference sub-tasks, while also providing trustworthy uncertainty quantification for better sensor placement and resilience against disturbances. AI
IMPACT This research could lead to more efficient and resilient traffic management systems in large cities.