Enhancing In-Car Navigation Systems with Personal Experience

Computers are extremely powerful for data processing, but less adept at handling problems that involve subjective reasoning. People, on the other hand, are very good at these kinds of problems. We present a framework for adding subjective human experience to in-car navigation systems. People often rely on their own experience when planning trips, choosing the route that seemed fastest in the past, the one that was the prettiest, or the one recommended by a friend. This led us to develop a set of methods to help people record their personal driving history, add rich annotations, and share their data with friends and family, or even the broader community. Users can then learn from their own data, or harness the multiplicity of individual experiences to enjoy new routes. This approach can be used in conjunction with traditional in-car navigation systems. INTRODUCTION Designers have just begun to realize the potential of using human power to accomplish tasks that computers, automated sensors and signal analysis techniques cannot do in a timely and accurate fashion. For example, researchers at Carnegie Mellon have created online games to attract users to annotate images or add facts to a knowledge-base [9], [10], [11]. Google Co-op invites people to manually annotate web pages to improve Google’s search capability. And search engines work so well in part because they analyze the result of human activity such as creating structured documents and linking among them. Yet, most current navigation solutions rely on objective, automatically collected data, ignoring human subjectivity. Personal experience plays a huge role in route choice. Traffic reports and websites do not always provide relevant, timely and accurate information and drivers often rely on their own knowledge – or impressions when planning trips to the airport, train station, friends, family, etc. Most use a trial and error approach [5]. People have different criteria for the “best” route: fastest, most scenic, lowest fuel consumption, or even perceived safety so combining objective and subjective measures may enhance the route selection process. Our approach differs from other route planning systems by starting with a decentralized the data collection process (Figure 1). By recording their own driving history, people can learn about the routes they actually use. By adding annotations to those routes, they will later benefit from data that cannot be recorded automatically, like the beauty of the scenery or the feelings of insecurity. Aggregating personal data helps discover patterns and confirm (or not) impressions drivers have about the characteristics of routes. While all users get immediate benefit from collecting and analyzing their personal data, some will reap greater benefits from sharing with friends and family, and then may choose to fully anonymize the aggregated data, share it with larger communities, and access more route data in return. Figure 1: Personal experience aggregation and sharing. Drivers record, annotate and aggregate their personal data, and can choose to share with friends, or a larger community. In this paper, we present a design framework for collecting, aggregating and sharing personal driving history as a case study of taking advantage of untapped human experience. We begin by discussing a survey we conducted to determine the type of questions people have when selecting routes, and how willing people are to share their personal driving experiences. We then discuss several scenarios using an early prototype to illustrate the recording and viewing of routes and mock-ups to show how annotations and share might take place. We also discuss methods of maintaining accuracy and anonymity as well as techniques for aggregating large collections of experiences and presenting that information to users in a meaningful way.