RouteLens Is Your Personal Route Planning Consultant
By Gabe Goldberg, HCIL Media Fellow
People pick driving routes in different ways. I work out a good route and stick with it; for new trips I'll check online maps or my fat road atlas. But I doubt that my wife ever drives the same route twice -- her spirit of adventure takes her down new streets trying to avoid just one more traffic light.
Each approach has advantages: I'm comfortable with familiar travel and my wife sometimes finds great new shortcuts. On the downside, I sometimes miss better routes, and my wife occasionally regrets excursions off known pathways. And we'll never agree on what routes are best, since she likes to keep moving even if it takes driving far out of the way and I don't mind a few traffic lights if I'm driving the shortest route. But we agree that traffic congestion is worsening and that we often lack good information for picking routes.
Wouldn't it be nice to have a personal route selector, getting advice based on data collected on previous trips? Current mapping and routing systems have disadvantages: First, being based on average or even best-case conditions, they're not always terribly accurate. Second, some local advisory systems aim more to improve traffic management than help individual drivers. So people often rely on personal experience when planning trips, even adding hunches and superstition to our mental models.
This sort of challenge fits the University of Maryland's Human Computer Interaction Laboratory (HCIL) mission: developing compelling solutions as foundations for continued innovation. RouteLens, a recent HCIL research project envisioned helping drivers choose best routes for common trips based on their personal driving histories.
Before building RouteLens, researchers Aaron Clamage, Ben Bederson, and Catherine Plaisant surveyed users regarding driving styles and preferences. Traffic conditions was by far the most important criteria for route selection, with time-of-day the next most considered factor. Interestingly, my wife's and my selection styles were nearly equally popular: "I have one way I like and I think it is the best route" vs. "I consider different routes and choose the one that seems best".
Radio is the most important information source, used nearly three times as often as any other factor. But TV, Internet, and friends/colleagues combined beat radio. Users rated predictable drive time as important, accepting longer expected trip duration in exchange for less likely delays.
Most desired functions were finding the best route for a given trip and time of day, suggesting the best time to leave for a given route, and being able to view traffic incidents along routes. Clamage was pleasantly surprised that more than 80% of users willingly shared data about their personal driving history with at least some people. Plaisant enjoys pooling travel information and advice with friends and colleagues, since they're likely to know her driving style better than a Web site might.
Focusing on drivers, researchers developed basic ingredients for their new approach. Rather than using sensor-derived data, drivers reported experiences during individual trips. Simultaneously, using software running on a cell phone or PDA (personal digital assistant) and wireless GPS (global positioning sensor) receiver, the system recorded trip aspects such as distance traveled, time of day, and speed.
Clamage notes that since less tangible factors also matter, drivers can annotate routes with other useful information such as scenic routes to seek or high-crime areas to avoid. And letting drivers pool personal driving histories builds a real-world database wiser than any of their individual experiences. Once drivers have taken enough trips, patterns reveal trends in travel time, stop frequency, number of stops, total stopped time, and (my wife's favorite!) speed.
RouteLens' desktop component provides the main visualization capability, used after importing recorded trip data files, to display routes and compare alternatives. It allows graphically analyzing individual trips and relating them to each other, permitting planning first-time trips to destinations near frequently visited destinations. Map views can zoom out for overview or in for detailed understanding and review.
Multiple routes can display simultaneously, graphically and with travel characteristics, for comparing alternatives. Displays can focus on locations or routes, depending on whether frequent or new destinations are being considered. Routes can be selected based on different characteristics such as expected travel time, maximum likely time, or total distance.
This project, sponsored in part by Microsoft Research, continues HCIL's tradition of understanding real-world problems, generating practical solutions framed within solid theoretical background.
So today's RouteLens is just the beginning. More extensive route annotation is required, along with more precise control over data sharing. History-based tasks will allow sharing route-based driving directions, comparing driver characteristics, and using fuzzy logic to find remembered places which can't quite be located. Finally, and most useful in real-time, will be adding visualization to RouteLens' mobile component so it's helpful en route.
