Simulations indicated that using a decision support tool to select alternative traffic control plans during non-recurring congestion in the Disney Land area of Anaheim, California could reduce travel time by 2 to 29 percent and decrease stop time by 15 to 56 percent.
Made Public Date


United States

Development and Evaluation of a Knowledge-Based System for Traffic Congestion Management and Control

Summary Information

This study investigated the use of a real-time knowledge-based decision support tool (Traffic Control Manager) to assist traffic operations center (TOC) personnel with selection of alternative traffic control plans after the occurrence of nonrecurring congestion. The ability of Traffic Control Manager to reduce delay during nonrecurring congestion was tested using the Dynasmart simulation model (Dynamic Network Assignment Simulation Model for Advanced Road Telematics) to measure average and total network travel time and stop time for scenarios "with" and "without" Traffic Control Manager.

The transportation sub-network input into the Dynasmart simulation model represented the Disney Land area of Anaheim, California. Actual traffic count data was input into the model to represent baseline saturation levels corresponding to high, medium, and low attendance levels at Disney Land. Each simulation had a two hour time horizon, and a total of 10 scenarios were run "with" and "without" the decision support tool. Scenarios included variations in attendance level, incident location (no incident, freeway, arterial), incident duration (10, 30, 60 minutes), capacity reduction (30 to 80 percent), and the changeable message sign (CMS) compliance rate (50 to 100 percent).

The Traffic Control Manager decision support tool was able to identify traffic congestion in the network by using a symbolic pattern matching process that mapped current network conditions to previously defined templates of typical traffic activity in the network. The template enabled baseline characteristics to be encoded into each link. If during the simulation there was a significant difference between baseline and current conditions, the Traffic Control Manager would predict the impact on local traffic flow and estimate the future demand in order to quantify a demand/capacity imbalance. At this point, the TOC operator would be notified to input additional information such as the length of any existing queues observed on closed circuit television (CCTV) and a best estimate of incident clearance time. Based on this input, Traffic Control Manager would select an optimum combination of control options using a predefined set of intersection signal control timing plans, freeway ramp meter strategies, and traffic diversion information on changeable message signs. All available control options were predefined by city traffic engineers prior to simulation. Based on heuristics, operators implemented the best solution. Once a selected solution was implemented, the Traffic Control Manager would estimate the time required for full recovery. With this information, TOC operators were able to monitor the network and identify any other incidents capable of interfering with the chosen solution. Throughout the recovery period, the Traffic Control Manager automatically tracked relationships between different congestion events. This enabled Traffic Control Manager to accurately estimate the source of any new congestion developing from other incidents, and recommend new countermeasures based on the potential impact of other solutions already implemented.


Simulations indicated the traffic diversion strategies and intersection signal control timing plan combinations chosen by the decision support tool reduced average travel time 1.9 to 29 percent, and reduced stop time 14.8 to 55.9 percent compared to scenarios without the tool.

Traffic Control Manager was less effective during high and low levels of saturation. Demand/capacity was harder to balance when the network was closer to saturation, and had limited effects when network demand was very small. Traffic Control Manager was most effective during medium levels of saturation.