Driver performance is generally quantified by the state of the vehicle relative to the local road and traffic environment. Unfortunately these vehicle-state-based metrics are limited in their diagnostic value when it comes to trying to assess how: (i) drivers individually adopted different control strategies, (ii) how they individually adapted to the issues under investigation (e.g., in-vehicle task execution, driver support system exposure, or impairment), or (iii) why drivers individually were more or less affected by the factor under study. By representing a driver's behavior in anidentifiable computational driver model, insight is gained into how drivers may differentially benefit or be impaired by the condition at hand. Such a model also shows how the myriad of possible performance metrics are all "necessarily" correlated. Based on test track car following data, a driver car following model is introduced and identified for each driver and used to show how drivers differ in their car following control strategies. It is demonstrated that the adopted target time headway (THW) strongly influences the associated control strategy (i.e., effort) as well as the safety margin (i.e., the minimum THWs experienced) and that subjects who adopt a longer target THW also exhibit a lower bandwidth control strategy (i.e.,less effort).
Abstract