【作者】 李爱娟；

【导师】 李舜酩；

【作者基本信息】 南京航空航天大学 ， 车辆工程， 2013， 博士

【摘要】 智能车辆是智能交通系统的重要组成部分，能显著提高行驶安全性和降低交通事故发生率，因而在军、民领域均具有重要的理论探索意义和工程实践价值。轨迹规划系统是智能车辆规划与控制系统的重要组成部分，与路径规划不同，路径规划生成的是静态无碰路径，而轨迹规划则是在路径规划的基础上考虑时间因素和车辆状态，不仅生成无碰路径还生成包括车辆行驶的各种状态参数和控制参数，如速度、加速度和行驶时间等。本文对智能车辆运动轨迹规划所涉及的相关内容：基于状态空间的轨迹规划方法，基于理性思维的自适应控制(Adaptive Control ofThought-Rational, ACT-R)认知模型的运动轨迹规划方法以及基于改进遗传算法的运动轨迹优化方法进行了较为系统和深入地研究。具体工作归纳如下：(1)考虑时间因素，建立了一种基于状态空间的曲率连续轨迹规划方法。为了能够生成适应于不同行驶环境且曲率连续的行驶轨迹，本文在最优控制理论的基础上进行了改进，并建立了多目标动态评价函数；通过不同权重对轨迹特征影响规律的研究，获得了权重调整的有效方法。仿真结果表明：本文提出的基于状态空间的轨迹规划方法能够针对不同道路环境动态生成满足多种约束条件曲率连续的行驶轨迹。(2)基于ACT-R认知模型，提出了一种考虑人的行为特性的轨迹规划方法。该方法将基于状态空间的轨迹规划方法和ACT-R认知模型结合起来，并以ACT-R认知模型为核心。首先由初始化模块生成初始权重集；然后用基于状态空间的轨迹规划方法生成轨迹，并将轨迹的特征值提取出来传给估计模块，在估计模块中对轨迹的特征值进行估计，判断轨迹的特征值是否满足约束条件，如果满足约束条件则将生成的轨迹返回给决策层；如果不满足，则通过权重调整模块对相关的权重进行调整，最终生成符合人的行为特性的行驶轨迹。仿真结果及模型车的试验验证表明：提出的方法能够生成具有人的行为特性，且满足多个约束条件的行驶轨迹，验证了本文方法的可行性。(3)根据车辆动力学约束条件，发展了一种将B样条曲线和改进的遗传算法相结合对运动轨迹进行优化的方法。首先利用B样条曲线曲率连续和局部支撑性的性质对生成的轨迹参数化，利用已知轨迹上型值点位置反求控制点，确定B样条曲线的形状；然后基于改进遗传算法对参数化轨迹进行多目标优化，来实现B样条曲线参数的确定。具体优化目标包括：每段B样条曲线的始末型值点位置，相应型值点位置的车辆速度、加速度和相邻型值点之间的时间间隔。仿真结果表明：优化后的轨迹较优化前的轨迹能够更好地满足动力学约束条件。(4)基于CarSim仿真软件，设计了一种对本文方法进行验证和分析的方法。在Matlab和ACT-R软件中由运动轨迹规划算法在采样周期内根据车辆将要行驶的环境信息对轨迹进行规划，并将规划的轨迹及轨迹的参数提供给CarSim理想控制器，用作动力学仿真软件中的路径和参数输入，理想控制器控制车辆模型在虚拟场景中沿着规划轨迹行驶，并将车辆行驶的轨迹和行驶过程中的各种响应参数反馈给运动轨迹规划算法，再由运动轨迹规划算法判断规划的轨迹及其参数是否满足约束条件，并对不满足约束条件的轨迹修正后重新输出轨迹，如此循环至仿真结束。分别进行了三组比较研究：本文方法和两种文献中路径规划方法的仿真比较；车辆以不同速度沿着规划出的稳态圆周轨迹、蛇形线轨迹和双移线轨迹行驶的仿真比较；以及仿真结果与文献实车试验结果的比较分析。结果表明：本文方法比文献路径规划方法更加有效可行，本文方法规划的车速比其它车速更能满足约束条件且仿真数据与文献实车试验数据吻合。综上所述，本文研究了智能车辆的运动轨迹规划方法，并结合仿真软件CarSim和模型车试验等方法对提出的运动轨迹规划方法的可行性和有效性进行了验证，其研究结果可以为后续跟踪控制器提供连续有效的运动轨迹，便于控制器实现对车辆的自动控制。更多还原

【Abstract】 Intelligent Vehicle (IV) is an important constituent of the Intelligent Transportation System (ITS).Intelligent Vehicle can improve the road traffic safety and reduce the traffic accidents remarkably, so ithas important theory searching significance and engineering practice value in military and civilterritory．The trajectory planning system is an important constituent of Intelligent Vehicle’s planningand control system. The trajectory planning is different from the path planning, the path planning onlygenerate the static collision-free path, while the trajectory planning not only generate collision-freepath but also relevant state parameters and control parameters, such as velocity, acceleration anddriving time, etc. The relevant contents of Intelligent Vehicle’s motion trajectory planning are studiedin this paper. The research contents include state space based trajectory planning method,ACT-R(Adaptive Control of Thought-Rational) based motion trajectory planning method and motiontrajectory optimized method. Concrete contents of the research have been summarized as follows:(1) Considering the time factor, a state space based trajectory planning method is presented. Inorder to generate curvature-continous and dynamic suitable different driving environment’ trajectory,based on the optimal control theory an improved trajectory planning method is presented in this paper.The multi-objective dynamic cost function is built in the improved method. The influence law ofdifferent weights to trajectory’s features is studied and the effective weights regulation method isobtained. The simulation results showed that the state spaced based trajectory planning method couldgenerate dynamically adaptive different road environments’ driving trajectory. The trajectory iscurvature-continous with several constraints are obtained.(2) Based on ACT-R cognitive model, a trajectory planning method with human’s behavioralcharacteristic is introduced. The state space based trajectory planning method and the ACT-Rcognitive model are contacted, the ACT-R cognitive model is the core of the method. Firstly, theinitialization module initializes the weight value set. Secondly, the trajectory is generated by the statespace based trajectory planning method and the trajectories’ feature values are extracted. Lastly, theevaluation module evaluates the trajectory’s feature values, make sure of the feature values underconstraints and return the trajectory. If the constraints are not obtained, the weight regulation moduleis used to regulate the weight value to generate the driving trajectory with human’s behavioralcharacteristic. The simulation results and model vehicle’s experiment results showed that the methodwas feasible. The driving trajectory generated by the method has human’s behavioral characteristicand several constraints are obtained.(3) On the basis of the vehicle’s dynamic constraints, a motion trajectory optimized method combined the B-spline curve and improved genetic algorithm is presented. First, the B-spline curve’sproperties of curvature-continous and local support character are used to parameterize the generatedtrajecroy. In order to make sure of the B-spline curve’s shape, the data point positions are known tocalculate the control points. Then, the improved genetic algorithm is used to multiobjective optimizethe parameterized trajectory. The parameters that can confirm the B-spline curve’s shape areoptimized to satisfy the dynamic constraints. The optimized objectives include data points and relativedate points’ velocity, acceleration and the time interval between the adjacent data points. Thesimulation results showed that the optimized trajectories’ dynamic constraints was obtained andoptimized trajectory was better than the unoptimized trajectory.(4) Based on the CarSim simulation software, a test method of this paper’s method is designed.The motion trajectory planning method generates the motion trajectory and the trajectories’parameters within the sampling period in the Matlab and ACT-R sofware. The generated trajectoryand trajectories’ parameters as CarSim’s path and parameter inputs are supplied to CarSim’s optimalcontroller. The optimal controller controls the vehicle model to drive on the planned trajectory invirtual scene simulation. The vehicle’s response parameters are extracted and reponsed to the motiontrajectory planning algorithm when the vehicle model driving on the planned trajectory. The trajectoryplanning algorithm estimates the parameters if under the constraints, if the trajectory is not under theconstraints, the trajectory planning algorithm modifies the trajectory and output the new trajectory.The cycle is done until the end of the simulation. Three group comparation studies were done to testthis paper’s method. Three couple comparation studies include the simulation comparation betweenthis paper’s method and two article’s path planning methods, the comparation of vehicle driving withdifferent velocities on the planned steady circular trajectory, serpentine curve trajectory and doubleshift line trajectory, the comparation between simulation results and other article’s experiment resultsof the real vehicle. The research results showed that this paper’s approach was more effective andfeasible than other article’s path planning method, the vehicle driving at the velocity planned by thispaper’s method could satisfy several constraints while the vehicle driving at other velocities could notsatisfy several constraints, the simulation datas were equal to the article’s real experiment datas.In summary, this paper focus on the sudy of Intelligent Vehicle’s motion trajectory planningmethod, the CarSim simulation software and the model vehicle’s experiment are used to test thevalidity and feasibility of the motion trajectory planning method. The studied results can providecurvature-continous and effective motion trajectory for the controller, so that the controller couldauto-control the vehicle driving on the road.更多还原