Research On Decision-support Model For Police Dispatching Based On Automatic Incident Detection In Urban Traffic

In the process of urban road traffic organization and management practices, the police dispatching decision in traffic incident management is an important task for police work efficiency. Capturing traffic incident timely and making high performance decision assure the validity and timeliness in traffic incident management, have vital role on reducing the loss caused by incident and relieving the work pressure of polices.This paper researches on above problems, presents an integrated solution of automatic incident detection (AID) and decision support for police dispatching. The innovations in this paper are following:1) build an improved AID model,2) based on this AID model, build a decision support model for police dispatching.Firstly, this paper researches on AID algorithms based on travel time. This paper improves the existing travel time measure method which based on vehicle length matching, presents a new AID algorithm. The algorithm obtains accurate matches and travel time by vehicle length match probabilities between upstream and downstream detectors and partial match density. The algorithm uses the standard normal deviate of travel time distribution to test known and unknown vehicle travel times, capture the abnormal travel time and detect traffic incident.Secondly, this paper researches and builds a decision support model for police dispatching based on Bayesian decision theory. The AID information is an input parameter of decision support model, hence the uncertainty in AID information is an importation factor in the precision of the model. Therefore, in order to reduce the uncertainty, the decision support model measures decision utility by total delay, combines the conclusion of incident state by AID algorithm and police, modifies the actual incident state probability under the lacks of clear description of incident, and ultimately control the risk implicated in decision-making errors.Finally, this paper designs a series of experiments, uses the actual traffic and incident data to verify the AID model and decision support model.