基于梯形规划曲线的智能车速度规划算法研究

Research on Intelligent Vehicle Speed Planning Algorithms Based on Trapezoidal Planning Curve

针对QP(Quadratic Programming)算法应用于智能车速度规划时,存在停车过程减速较晚而导致减速距离短、平稳性差的问题,文中提出了基于梯形规划曲线的智能车速度规划算法.首先建立速度规划的 QP模型并求解;然后分析不同初速度下基于梯形规划曲线的停车过程,将其结果作为非线性约束来实例化QP模型并再次求解;最后通过仿真实验和实车实验对比分析 QP算法和所提算法的实验结果.仿真实验中,分别以39.8km/h,31.5km/h,20.6km/h的初速度进入停车过程,速度变化曲线表明所提算法能够将减速开始时间提前,初步表明该算法具有优化效果;实车实验中,较QP算法而言,所提算法将3种初速度下的停车过程分别提前5.9s,5.0s,3.7s,平均加速度绝对值分别减少0.5m/s^2,0.5m/s^2,0.4m/s^2,最大加速度绝对值分别减少0.16m/s^2,0.33m/s^2,0.35m/s^2.仿真实验和实车实验表明,所提算法的改进效果明显,具有显著的优化作用.

Aiming at the problem of short deceleration distance and poor stationarity caused by late deceleration in par- king process when QP (quadratic programming) algorithm is applied to speed planning of intelligent vehicles,there are some problems such as short deceleration distance and poor stationarity caused by late deceleration in the stopping process.This paper presented an intelligent vehicle speed planning algorithm based on the trapezoidal programming curve.Firstly,the QP model of speed planning is established and solved.Then,the stopping process based on trapezoidal programming curve at different initial speeds is analyzed,and its results are considered as nonlinear constraint to instantiate and solved QP model .Finally,the experimental results of QP algorithm and the algorithm were compared and analyzed through simulation experiment and real car experiment.In the simulation experiment,the initial speed of 39.8 km/h, 31.5 km/h and 20.6 km/h was used to enter the parking process respectively.The speed curve shows that the proposed algorithm can advance the deceleration time,which preliminarily shows that the algorithm has the optimization effect.In the real vehicle experiment,compared with QP algorithm,the proposed algorithm advances the parking process of the three initial speed by 5.9 s,5.0 s and 3.7 s,the absolute value of the average acceleration decreases by 0.5 m/s^ 2,0.5 m/s^ 2 and 0.4 m/s ^2 ,the absolute value of the maximum acceleration decreases respectively by 0.16 m/s ^2, 0.33 m/s^ 2 and 0.35 m/s^ 2 .The simulation and real vehicle experiments show that the improved method has obvious improvement effect and significant optimization effect.