Personalization for Route Guidance


Most automobile navigation systems provide a route that minimizes the sum of the costs on its individual road segments, with the cost for each segment depending on its length or estimated driving time. However, the presence of digital maps lets one access other attributes of routes, such as the number of turns and intersections, the types of segments composing them, and even whether the driver has followed the route before. Combining such attributes gives a more flexible cost metric, but the relative importance of each feature can differ across drivers, and specifying this information manually is a tedious process.

This project aims to develop an adaptive route advisor that develops a model of a driver's preferences over time. For each navigation task, the system presents a number of alternative routes generated using different cost functions. From these, the driver can select a route or request a new route that improves one of them along a given dimension. When the driver finally decides on a route, the advisor records its features and those of the alternatives, then uses machine learning to revise its model of the driver's cost metric, which it uses in generating routes for the next navigation task. Preliminary experiments indicate that drivers differ considerably in their route preferences, and that our approach can learn to accurately predict these differences.

This research was funded by DaimlerChrysler Research and Technology.


Contributors to the Project

  • Dr. Claude-Nicolas Fiechter

  • Brian Johnson

  • Dr. Simon Handley

  • Professor Pat Langley

  • Annabel Liu

  • Dr. Seth Rogers

  • Daniel Russakoff

  • Related Papers

    Fiechter, C.-N. & Rogers, S. (2000). Learning subjective functions with large margins. Proceedings of the Seventeenth International Conference on Machine Learning (pp. 287-294). Stanford, CA.

    Rogers, S., Fiechter, C.-N., Thompson, C. (2000). Adaptive user interfaces for automotive environments. Proceedings of the IEEE Intelligent Vehicles Symposium. Dearborn, MI.

    Rogers, S., Fiechter, C., & Langley, P. (1999). An adaptive interactive agent for route advice. Proceedings of the Third International Conference on Autonomous Agents (pp. 198-205). Seattle: ACM Press.

    Rogers, S., Fiechter, C., & Langley, P. (1999). A route advice agent that models driver preferences. Proceedings of the AAAI Spring Symposium on Agents with Adjustable Autonomy. Stanford, CA: AAAI Press.

    Rogers, S., & Fiechter, C. (1998). An adaptive interactive agent for route advice. Unpublished manuscript, Daimler-Benz Research & Technology Center, Palo Alto, CA.

    Rogers, S., & Langley, P. (1998). Personalized driving route recommendations. Proceedings of the AAAI-98 Workshop on Recommender Systems. Madison, WI: AAAI Press.

    Rogers, S., & Langley, P. (1998). Interactive refinement of route preferences for driving. Proceedings of the AAAI Spring Symposium on Interactive and Mixed-Initiative Decision-Theoretic Systems (pp. 109-113). Stanford, CA: AAAI Press.

    Rogers, S., Langley, P., Johnson, B., & Liu, A. (1997). Personalization of the automotive information environment. Proceedings of the Workshop on Machine Learning in the Real World: Methodological Aspects and Implications. Nashville, TN.


    For more information, please send email to rogers@rtna.daimlerbenz.com .