A Research On Longitudinal Takeover Strategy Of Driving Right Considering Driver Characteristics

Autonomous vehicle has received extensive attention in the automotive industry,for its significant advantages in improving driving safety and lightening the burden on drivers.However,due to various limitations of technology and social conditions,the humanmachine cooperative vehicle will still exist for a long time in the future.When detects that the current driving conditions exceed its design limit,such as missing lane lines,hardware failures,bad weather and other unexpected situations,the automated control system will send driver a takeover request of driving right.Hence,a safe and reliable takeover strategy of driving right is an important basis for the realization of humanmachine cooperative driving.For the driving risk assessment algorithm,the existing takeover strategy miss the impact of inherent coupling of between risk indexes when design the risk domain from single dimension,and thus fails to reflect real driver’s risk perception characteristics.In addition,they assess the takeover risk without the consideration of the driver specificity and takeover time characteristics.Since the takeover strategy cannot be updated adaptively according to driver characteristics,it cannot be applied to different drivers and takeover scenarios.In order to solve these problems,this paper proposes a longitudinal take-over strategy based on driver characteristics.According to the conditions of safe takeover of driving right,the overall scheme of takeover strategy of driving right is determined,with several key issues of the scheme are discussed,and corresponding solutions are put forward.Including the design of Driving Risk Acceptance Set(DRAS),the autoadaptability of take-over strategy to driver specificity,and the dynamic auto-adaptability of take-over strategy to driver takeover time.The specific research are as follows.Firstly,on basis of the analysis of drivers’ risk perception characteristics,the longitudinal driving risk acceptance domain is designed.To assess the driving risk reasonably,it is necessary to analyze the driver’s ability to accept the risk.This paper obtained the risk thresholds of each risk index based on the statistics of natural driving data,and used the thresholds to evaluate driving risks.Finding the risk assessment method using a single Time to Collision(TTC)or Time Headway(THW)index could not correctly identify the driving risk in some conditions.Meanwhile,correlation analysis was used to verify the coupling characteristics between the risk indexes.In order to design a DRAS confirms to drivers’ risk cognition characteristics,a method of constructing DRAS from two-dimensional space based on TTC and THW indicators is proposed.Avoiding the disadvantage that a single-index cannot correctly identify driving risk in some conditions,and improving the influence of coupling characteristics between indicators on risk assessment.The driving risk model was established by Gaussian mixture model,and the boundary expression of DRAS was derived by cumulative distribution inverse function.After the design of DRAS was completed,the rationality of DRAS was verified by a significance test on the DRAS and risk domain witch ignored the coupling characteristics between indicators.Furthermore,to make the takeover strategy adapted to different drivers and takeover scenarios,an analysis about the adaptability of the takeover strategy to driver characteristic is conducted.On the one hand,to satisfy the driver specificity,it is inappropriate to use a unified risk set for assessment.Thus a recursive algorithm-based online update algorithm of the risk set boundary is designed,which can update the risk acceptance set adaptively according to the driving data of different drivers.On the other hand,taking into account the characteristics of the driver’s takeover time changes dynamically with the state of “human-vehicle-road”,the take-over risk should be judged by the driving risk at each moment from the current to the completion of take-over.The parameter of takeover time is introduced into the takeover risk assessment algorithm,which is estimated in real time according to the state of “driver-vehicle-road”.Predicting the system state during the takeover time to estimate the change of driving risk,then the adaptive of the takeover strategy to the driver’s takeover time is realized.Finally,the effectiveness of the proposed longitudinal takeover strategy of driving right is verified based on the driving simulation platform.The results show the takeover strategy can assess the takeover risk timely and reasonably with different driver and takeover scenarios,effectively improve the takeover safety of the human-machine cooperative vehicle.