Adjustment of an Assisted Driving Function Based on Driver-Initiated Takeovers

When using advanced driver assistance systems, drivers often prefer a slightly different driving style than the one set by the system. They express their personal preference by intervening in the active system to correct its behavior. This paper proposes a framework which uses the information contained in these driver-initiated takeovers to derive needed adjustments to a predictive longitudinal driving function on repeatedly traveled routes. Two dimensions of possible adjustments are analyzed: the need for driver individualization and the need for local map adjustments at specific locations. When applying the framework to a naturalistic driving dataset, the results show that individual drivers show differing takeover behavior which proves the need for individualization. Furthermore, drivers' behavior varies by location, so incorporating the location dimension enables optimizations which could not be achieved with general function parameter adjustments alone. The analysis of driver behavior on the location level reveals that drivers consistently intervene in 24.9 % of unique locations, indicating a need for individualized map updates. ... mehr65.8 % of locations require no adjustments as drivers do not intervene there, and in 9.4 % of unique locations, drivers show inconsistent behavior, making it impossible to judge the need for a local adjustment.