The number of miles driven in the United States doubled from 1970 to 1990 and is expected to double again by 2020. Roadway congestion costs $43 billion per year in lost productivity and fuel costs for the 50 largest cities in the United States (source Federal Highway Administration 1996). As it becomes impossible to build additional roads to meet the demand, more effective management of our existing roads is necessary.

To meet this challenge many state and city transportation departments are developing traffic management centers (TMC). Trained operators in the TMC's oversee the roadway network and respond to incidents. TMC operators use a variety of resources to respond to roadway incidents including: changeable message signs, ramp metering, tow trucks, and highway advisory radio. Using these resources effectively requires experience and knowledge. Making the right decisions can keep a roadway incident from becoming a traffic nightmare.

Knowledge Systems Design has developed a knowledge base that captures incident response knowledge. The knowledge base can be integrated into existing information systems within a TMC. This knowledge base captures critical knowledge that experienced traffic operations managers use when making incident response decisions. When an incident is detected in the roadway system the knowledge base is invoked and analyzes the current situation. Based on the analysis, response recommendations are generated for the operators. The TMC operators can choose to implement some or all of the incident response recommendations.

Knowledge Base

The knowledge base uses both rules and model-based representations to capture the incident response knowledge. Rules capture conditions under which different responses are applicable. For example, if an incident occurs during peak traffic hours and the clearance time of the incident will be more than 30 minutes, then sending a traffic alert to local radio stations is one of the recommended responses. Determining the level of traffic impact of a given incident is knowledge that is also captured in rules.

When reasoning about a complex system, such as a roadway network, understanding the relationships between parts of the system is critical. The KSD knowledge base uses a model of the roadway network to reason about the location of incidents. Using the network model, diversion paths around the incident can be analyzed to assess their effectiveness. The model can also be used to search upstream of the incident for changeable message signs CMS (i.e. motorist information signs). Messages for each of the signs are constructed based on the location of the sign relative to the incident. Diversions and messages for CMS are additional incident responses that may be recommended based on the incident situation.

Since response knowledge varies for each traffic management center, KSD personnel work with department of transportation employees to capture incident response knowledge specific to their district. Each roadway network is different and thus the set of possible responses vary based on the resources available in each area and the local traffic patterns. Adding this district specific knowledge is not difficult since the basic structure of the KSD roadway incident management knowledge base is pre-built.

Integrating with Existing Systems

The roadway incident management knowledge base can be integrated with existing TMC information systems. Off-the-shelf bridges are available for most standard databases (Oracle, Sybase, and ODBC compliant databases). Custom bridges can be written to existing data collection systems such as roadway speed and volume measurement sensors. The knowledge base can be integrated with many types of user interfaces, from complex geographic information displays to simple text based terminals.

Key Benefits

• The best decision making knowledge is available to all TMC operators.
• The knowledge base quickly assesses many factors when making recommendations, this ensures a critical detail is not overlooked.
• Legal guidelines can be made part of the knowledge base to ensure that specific procedures are followed for certain incident types.
• Incident response recommendations are available within seconds of initial incident detection. During peak traffic periods every minute of reduced congestion is a big payoff.
• Records are kept of the response decisions made for all incidents.