Study On Forward Collision Warning Method Adaptive To Driver Characteristics

Driver behavior is the main cause of traffic accidents. Forward Collision WarningSystem can enhance driving safety by providing acousto-optic warning signal to driverswhen it detects collision threats. However, high false alarm rate and poor acceptabilityproblems still challenge current FCW as the result of the complexity of roadenvironment, diversity of driving conditions and variability of driving conditions. Inorder to improve FCW’s adaptability to the volatility and individual difference ofdrivers, a forward collision warning method adaptive to driver characteristics isproposed. On the basis of the analyzing result of driver’s multi-lane vehicle followingbehavior, combining with adjacent lane vehicles’ cut-in intention and relevance offollowing status, the proposed algorithm realizes comprehensive assessment of riskfrom multi-lane preceding vehicles by car-following parameter amendment. Thestrategy of FCW adaptive to driver characteristics is designed, and its performance isverifie.Based on pointing out the limitation oftraditional following behavioranalysis,multi-lane vehicle following behavior analysis is carried out. The concept ofgrey relational degree is introduced to describe ego vehicle’s relational behavior topreceding vehicle in-lane, left lane preceding vehicle and right lane preceding vehicle.In order to support the amendment of behavior parameters, the rules of ego vehicle’sbehavior in multi-lane following condition is drawed.To design the parameter amending method considering multi preceding vehicles’traveling status, a dynamic programming law of multi-target recognition method isestablished. The tracking difficulty of lateral position of adjacent lane vehicle basedon vehicular radar is analyzed, and a cut-in risk identification method based on TotalLeast Squares is proposed. Take the adjacent lane vehicles’ cut-in intention and theirrelational following behavior into consideration, a fuzzy-logic method is used todetermine three preceding vehicles’ respective weight in multi-lane driving. Also, thecar-following parameters are matched and amended according to the driving conditionin multi-lane road.Traditional FCW algorithm conception with fixed model or fixed warning threshold is broke. The amended car-following parameters are used as system input,and the algorithm of adaptive FCW is designed. The online identification method ofbehavior model is established, and model parameter can be adjusted online. Thewarning threshold prediction method and optimal threshold search method areestablished, and the warning threshold can also be adjusted online. The logic of modelparameter adjustment is designed.Offline simulation of real car test data is adopted to verify the effectiveness of theFCW system. Four algorithm patterns are compared through the statistics ofperformance index and the performance trends during online usage. The results showthat, adaptive FCW algorithm can automatically adapt to drive’s volatility andindividual difference, and it performs best in four patterns. Also, as the action timeincreases, the adaptability of the system gradually improves and false alarm ratedecreases.