University of Waterloo
Engineering 5 (E5), 6th Floor
Phone: 519-888-4567 ext.32600
Design Team: J. A. Morton
Supervisor: Professor L. Fu
The traffic in major North American metropolitan centers is increasing yearly. People want to be able to find their way from one place to another. Whether they are just visiting a city or are long-time citizens, an escort familiar with the city’s layout and possessing accurate, up-to-the-moment traffic details would be invaluable. Modern automobiles and Internet services do provide static maps giving driving directions from place to place. These can be modified to accommodate a very limited range of user preferences (e.g. avoidance of toll-roads). The relative ease of traffic information dissemination through the radio is often delayed, sporadic, and sometimes inaccurate. The user might like to find the route to their destination with the least number of traffic lights, or fewest turns. Perhaps they prefer to avoid the highway. Or maybe they are searching for the fastest route based on current traffic conditions. Maybe they are less concerned with time and prefer the scenic and safe route. Perhaps the driver wishes to travel along a specific street to get a coffee. These scenarios cannot be handled by today’s existing retail routing information systems.
This workshop will address the needs of the average person who travels within a city primarily by car and occasionally via public transportation. The objective of this proposed workshop is to develop a route selection mechanism that can adapt to a user’s preferences. The system would present the user with a series of route choices and the user would then pick one. The system would add the user’s choice to its learning process to find routes possessing similar characteristics. The next time the driver asks for directions, the choices they are presented with will more closely resemble their choices from the past.
The lack of reliable or dynamic traffic information makes wayfinding difficult to all but the most experienced motorist. The problem with current routing models is they are static. An adaptive route selection model would adjust the routes found from origin to destination to any driver’s particular style. This model must be able to accept real-time traffic data, user input, and environment data. It should adapt (within a given threshold) to any user given a minimum of training data. The feasibility of such an endeavour may be limited due to computing resources, the existence of accurate data, and financial considerations. Implementation of the model will take place either on an in-vehicle navigation system or personal digital assistant (e.g. iPaq™ Blackberry™). This implementation will be tested with limited geographic data and no Global Positioning System (GPS) functionality. Users of various types will operate the model, and performance will be measured on the number of iterations the model takes to accurately predict the user’s route selection habits. It will also be appraised on the “closeness of fit” to a given user type. The practicality and possibility of implementation of such a model will then be extrapolated into the “real world.”