A vehicle dynamics based algorithm for driver evaluation

Abstract

Driver evaluation and performance monitoring is crucial for safety and risk assessment. This paper proposes a novel algorithm for evaluating driver performance using vehicle state data. This indirect method of evaluation has the advantage of being unobtrusive and cheap as it uses only the easily accessible real time vehicle data. The proposed algorithm can eliminate subjectivity and automate the evaluation process. The driver being evaluated is asked to maintain lane center and drive smoothly while steering input, pedal position and acceleration data is collected. Steering activity is evaluated by defining an optimal input for lane keeping and comparing driver's input. Wheel slip and cornering conditions are considered for pedal activity evaluation. Vehicle acceleration data is used for estimating passenger comfort. An instantaneous and average index of performance is generated in real time as a final evaluation metric combining the three evaluation criteria. The Algorithm was validated using a vehicle simulator and the results are discussed.