凸近似避障及采样区优化的智能车辆轨迹规划.pdfVIP

3482024，60（1）ComputerEngineeringandApplications计算机工程与应用

凸近似避障及采样区优化的智能车辆轨迹规划

张宜旭，田国富，王海涛

沈阳工业大学机械工程学院，沈阳110870

摘要：针对结构化道路下作匀速运动的智能车辆避障轨迹规划问题，提出一种基于凸近似避障原理及采样区域优

化的智能车辆轨迹规划方法。引入凸近似避障原理，得到轨迹可行域范围；将采样区域分为静态采样区、动态采样

区两部分，并根据障碍物运动状态，另外划分动态、静态障碍物采样区；采用“动态规划（DP）+二次规划（QP）”思想求

解轨迹：利用五次多项式对采样点依次连接，建立动态规划代价函数并筛选得到粗略轨迹；通过二次规划及约束条

件的构造，对粗略轨迹进行平滑，最终得到最优轨迹。仿真结果表明：对于静态、低速、动态三种障碍物，该车能够有

效地得到平滑轨迹并避开障碍物。

关键词：轨迹规划；Frenet坐标系；凸近似避障原理；动态规划；二次规划

文献标志码：A中图分类号：TP242doi：10.3778/j.issn.1002-8331.2209-0003

TrajectoryPlanningMethodofIntelligentVehicleBasedonConvexApproximateObstacle

AvoidancePrincipleandSamplingAreaOptimization

ZHANGYixu,TIANGuofu,WANGHaitao

CollegeofMechanicalEngineering,ShenyangUniversityofTechnology,Shenyang110870,China

Abstract:Aimingattheproblemofobstacleavoidancetrajectorytrackingforintelligentvehiclesmovingatconstant

speedonstructuredroads,atrajectorytrackingmethodofintelligentvehiclesbasedonconvexapproximateobstacle

avoidanceprincipleandsamplingareaoptimizationisproposed.Theconvexapproximateobstacleavoidanceprincipleis

introducedtoobtainthefeasiblerangeoftrajectory.Thesamplingareaisdividedintostaticsamplingareaanddynamic

samplingarea,accordingtothemotionstateoftheobstacle,thedynamicobstacleandstaticobstaclesamplingareaare

alsodivided.Theideaof“dynamicprogramming(DP)+quadraticprogramming(QP)”isusedtosolvethetrajectory.The

samplingpointsareconnectedsuccessivelybythequinticpolynomial,andthedynamicprogrammingcostfunctionis

establishedandtheroughtrajectoryisobtainedbysearching.Throughthequadraticprogrammingandtheconstructionof

constraints,theroughtrajectoryissmoothedandtheoptimaltrajectoryisobtained.Thesimulationresultsshowthat:the

vehiclecaneffectivelyobt