Chung Hee Hwang, Noel Massey, Bradford W. Miller, and Kari Torkkola
We report on our on-going effort to build an adaptive driver support system, Driver AdvocateTM, merging various AI techniques, in particular, agents, ontology, production systems and machine learning technologies. The goal of DA is to help drivers have a safer, more enjoyable, and more productive driving experience, by managing their attention and workload. This paper describes the overall architecture of the DA system, focusing on how we integrate agent and machine learning technologies to make it support the driver intelligently and unobtrusively. The architecture has been partially implemented in a prototype system built upon a high-fidelity driving simulator, letting us run human experiments. The human driving data collected from the simulator are used as input to machine learning tools to make DA learn to adapt to the unique driving behavior of each driver. Once the DA demonstrates the desired capabilities, it will be tested in a real car in an actual driving environment.