Analysis of AEB False Positives in Urban Bus Operations

Abstract

False positive (FP) activations of Automatic Emergency Braking (AEB) in urban buses can cause unnecessary harsh braking and reduce system acceptance for situations involving both passengers and vulnerable road users. This report implements a traceable, criteria-based pipeline to identify candidate FP events from raw ASAM MDF 4 (MF4) recordings, structured as file-level screening, qualified brake-event detection, and FP classification. The implementation uses core Controller Area Network (CAN) signals to detect AEB-active braking events and applies collisionrisk and driver-response criteria for classification, where time-to-collision (TTC) is computed from recorded signals and used as one decision input. The TTC computation matches a ground-truth TTC reference provided by a colleague within the Group Trucks Technology team on the evaluated data, and the end-to-end detection pipeline runs on an internal dataset to automatically extract candidate FP events for further attribution and performance analysis.