Evaluating vehicle stability support systems by measuring, analyzing, and modeling driver behavior

Abstract: This thesis presents an investigation of near-accident behavior of truck drivers, with and without support from an electronic stability control (ESC) system. A critical scenario, involving both collision avoidance and vehicle stabilization on a low-friction surface, was studied in a driving simulator. The simulator experiment included a novel experimental paradigm, in which several measurements of critical maneuvering were generated per test subject. In this paradigm, ESC was found to provide statistically significant reductions of skidding and control loss, and the drivers were found to employ similar strategies for steering control as when they experienced the same scenario unexpectedly. These findings imply that the system should provide stability improvements also in unexpected maneuvering, something that has not been previously demonstrated for heavy truck ESC. A review of existing driver behavior models that can be used in simulation-based testing of active safety systems (such as, for example, ESC) is also presented. The review showed that, while a wide range of models has been proposed, the generated behavior can sometimes be more similar between models than what the model equations may suggest. Validation of models on actual near-accident behavior of real drivers has so far been very limited. Here, it is shown that an existing model of steering can reproduce the stabilization steering behavior observed in the simulator study. It is also demonstrated how this model can be mathematically linked to vehicle dynamics concepts, increasing its usefulness in applied contexts.