Identifying abnormal traffic congestion on non-signalised urban roads using journey time estimation

摘要

This paper describes a technique for estimating vehicle journey times on non-signalised roads using 250-ms digital loop-occupancy data produced by single inductive loop detectors. The technique was assessed to see whether abnormal periods of traffic congestion (caused by accidents and special events) could be identified using the journey time estimates produced along a key urban corridor in the city of Southampton. The technique used a neural network approach to provide historical journey time estimates every 30-seconds based on the average loop-occupancy time per vehicle (ALOTPV) data collected from the detectors during the previous 30-second period. Results showed that using the output from 8 detectors over 1149m, journey time estimates with a mean absolute percentage deviation from the mean measured speed (MAPD) of 15% were returned. These were achieved using a neural network trained on 7 days of morning peak period data. The journey time estimates produced were presented to the control room operator in the form of a moving graph, updating every 30-seconds. Results showed that the journey timeestimates identified 73% of the logged incidents on the test network during the analysis period.