Research On Target Object Selection For ACC System Based On Analysis Of Vehicle Trajectories

The development of ACC (Adaptive Cruise Control) first time made it possiblethat part of drivers operation could be fulfilled by an automatic system. Apart fromfunctions of conventional Cruise Control, ACC could recognize if there exists a validtarget vehicle via environmental information captured by frontal radar, and could alsokeep a certain distance to the target vehicle, which makes it an electronic system ofboth safety and comfort.Besides the utilization rate and the cost problem, target selection is a specialchallenge of ACC system. Since the driving trajectories change as the curvature oflanes change and preceding cars and the host car are always located at places ofdifferent curvature, it is necessary that the system makes the prediction of frontcurvature, thus to judge whether there exists a valid car and which is the valid caramong the cars found by the radar, also which is made difficult to accomplish with theuncertainty of each driver s operations. As for the prediction of curvature, somemethods and theories had already been presented within recent years, naturally thetheoretical basis of which is information of front curvature is contained in trajectoriesof preceding cars, so the point is how to read the information from trajectories. Alsothe necessity of differentiating drivers operations, especial the differentiation oflane-change and curve-entry/exit was proposed by some theories, which could helppredicting the curvature more precisely.It is believed in this paper that the target selection method should focus more onthe operations of drivers via analysis of preceding and host cars trajectories, and theclassification driving operations. Sequentially and obviously with the different kind ofdriving operations, different data processing method could be used to calculate frontcurvature very precisely with obtained trajectory data. Three diver s operatingconditions are defined in the paper for this purpose, which are in straight line (inS), on curve (onC) and in lane-change (inL), and the reorganizing method of operatingcondition of a certain car (preceding or host car) is designed as two steps: thepre-judgment and the judgment. Since most parts of a highway are straight or nearly straight, the pre-judgment designed for simplifying the computation and it s used tojudge that if driving operations the a driver makes demonstrate obvious intention tochange lane or enter a curve, which is called a active steer and if a active steer appears, the judgment will act as the second step to further classify operatingcondition, which mainly using methods of curve fitting and paired t-test.Besides, the pre-judgment and the judgment designed about operating conditionin the paper are both based on analysis of vehicle trajectories in a certain coordinate,which should be established at the first, and in fact, the variation of coordinatechanges the coordinate figures of points which constitute trajectories. Thus, a series ofupdating rules of the coordinates is set to guarantee the intermittent update, whichcould eliminate the accumulation of error to a great degree.After the confirmation of operation condition of preceding and host cars, detailedcalculating and recording methods for predictive curvature are introduced in the paper,in which if there exist more than one preceding car, weighting method would be usedto optimize the calculation, and if the deficiency of curvature occurs, certainpredictive formulas would be used to make a rough prediction. After all, the intentionof calculating curvature is to calculate the relative position between the host car and acertain preceding car (the lateral distance between preceding car and predicted centerline and the distance between the two vehicles alone the road s direction), which arealso gradually introduced in the paper.In the4thchapter, simulating method and environment, together with part ofsimulating results are introduced and presented, which adopts the simulation ofsingle-driver-in-the-loop, and employs Carsim, Simulink and hardware-in-the-looptest system dSPACE Simulator to fulfill the collection of driver input, environmentalsimulation, algorithm simulation and the simulation of radar, etc., and the adoptedmethod guarantee the repetitive analysis of a certain trajectory. The simulating resultsshow that the basic theories of this target object selection method for ACC systembased on analysis of vehicle trajectories are fundamentally correct, and thecomputational accuracy of it could satisfy the demand of ACC system under generalconditions.