An Optimal Longitudinal Control Algorithm For Automated Driving Considering Personalized Risk Characteristic

With the contradiction between the drastically increasing number of motor vehicles and road capacity limitation,in order to break through the bottleneck of road traffic development,the automated vehicle(AV)of the intelligent transportation system(ITS)has become the mainstream research direction of next generation traffic.Among all vehicle automation functions,longitudinal control,which enables AV to autonomously adjust its speed to maintain a proper distance with the preceding vehicle and further to improve capacity and enhance traffic safety,is of vital importance.As is known to all,different drivers tend to have heterogeneous driving styles.However,how to integrate personal driving behavior preference into AV’s longitudinal control has always been the key point and challenge in research.Based on the abovementioned background,considering the personalization of driver’s preference on stochastic disturbance and control comfort,a personalized stochastic optimal adaptive cruise control(ACC)algorithm for AVs incorporating driving behavior preference is proposed.Specifically,the state-space formulation is established with mixed disturbances stemming from system and measurement.Then the proposed controller is designed as a linear exponential-ofquadratic Gaussian(LEQG)problem,which utilizes the stochastic optimal control mechanism to feedback the deviation from the design car-following target.Meanwhile,the state feedback control strategy and output feedback control strategy are proposed for different application scenarios.By classifying driver’s preference for stochastic disturbance and by adjusting the values of the weight matrices in the cost function for different control modes,heterogeneous driving behaviors are generated.The influence of feedback control strategy,magnitudes of mixed uncertainties,and control modes on the performance of personalized automated driving are compared and analyzed based on simulation experiments.The results validate that the proposed approach can characterize different driving behaviors and its effectiveness in terms of reducing the deviation from equilibrium state.Furthermore,a surrogate safety measure based active safety evaluation method for AV’s longitudinal control is presented,and typical carfollowing scenarios are selected for algorithm’s comprehensive safety analysis.The analysis confirms that the proposed algorithm has excellent safety performance with the realization of personalized control.