Study On Driver Capability Based Take-over Strategy Of Cooperative Driving System

Cooperative driving system has been treated as an alternative solution as the system can allow human pilot and machine pilot in the control loop concurrently to exploit the advantages of the two subjects.Existing systems switch the control authority dependent on driver state and driving risk.However,driver’s inattentiveness does not mean that the driver is unable to continue with driving task and driver is likely to correct driving risk by himself.Thus,they are not suitable for designing the switching strategy.To solve the problem,a switching strategy based on driver capability evaluation is proposed.First of all,the feasibility of the takeover strategy based on driver’s status and driving risk assessment is studied in this paper through experiment.By analyzing the relationship between different driver’s status and the number of dangerous incidents and the situation of driver’s self-correction of risk,the drawbacks of under-intervention and overintervention based on the existing take-over strategy are found,which proves the importance to improve the strategy.Then,the general idea of switching strategy referring to driver capability is presented and the method to evaluate driver capability using unreliable domain of driver capability(URDC)is introduced.As well,we design the indicators to calculate the bounds of URDC,which include driving risk and driver correcting capability.Driver model is applied to design the indicator to evaluate driver correcting capability so we compare different types of driver models including mechanism model and marching learning model to determine which one to use in our research.After that,the method to compute the bounds of URDC is designed and we managed to make URDC update online and in real time to ensure it adapted to the driver characters.The lognormal distribution is used to describe the characteristics of URDC indicators and the inverse of cumulative distribution function is used to calculate the boundary value of URDC.The mean and standard deviation of distribution are updated in real time by recursive method to realize the self-learning function of boundary value.Finally,the validity of the takeover strategy proposed in this study is evaluated by the data collected from the simulated driving.The results show that,compared with the existing strategies,the switching strategy based on driving ability assessment effectively reduces the under-intervention and over-intervention.