【作者】 王翔；

【导师】 高振海；

【作者基本信息】 吉林大学 ， 车辆工程， 2020， 硕士

【摘要】 在传统的“人-车-路”交通系统中,人是最薄弱的环节,也是造成交通事故的主要原因。在自动驾驶环境中,由于部分或全部替换了驾驶人环节,所以自动驾驶技术被认为可以极大地解决交通安全问题。在自动驾驶技术的众多应用场景中,园区场景由于速度较低、工况较为简单而受到众多企业和高校的关注,是自动驾驶可率先实现的场景之一,园区自动驾驶成为近几年的研究热点。本文依托科技部重点专项“电动自动驾驶汽车关键技术研究及示范运行”,在园区场景内研究了局部规划与跟踪控制技术,并基于仿真和实车平台进行了算法验证。针对局部规划问题,首先本文引入了弗莱纳框架,将局部规划问题分解为两个一维问题,降低了规划难度。随后考虑到自动驾驶车辆行车环境动态变化,本文对交通环境的行车风险进行了分析,并基于势场法对行车环境进行建模,通过引入自适应调节因子,实现动态障碍物的风险影响范围随相对速度变化而自动调节。最后通过五次多项式曲线实现候选轨迹的生成,本文通过对状态空间中的目标状态进行合理采样,分别构建横纵向轨迹簇并进行组合处理,作为候选集合。本文设计了考虑安全性、舒适性和高效性的多目标评价函数,来对候选轨迹进行评价决策,筛选出本周期内最优轨迹。在下一时刻重复以上步骤,如此滚动进行,以实现局部规划的动态更新。针对轨迹跟踪问题,本文采用解耦控制策略分别构建了车辆横纵向控制器。首先基于模型预测控制理论构建了车辆横向控制器,采用车辆横向动力学模型作为预测模型,设计控制量和输出量约束条件,建立MPC优化问题,并通过将其转化为二次规划问题进行求解,该方法兼顾了跟踪的准确性与舒适性。对于纵向控制,本文依据最优预瞄理论设计了上位控制器,实现期望加速度决策,通过车辆纵向逆动力学模型建立下位控制器,实现加速度到车辆输入的转换,从而实现了分层式纵向控制策略。最后,本文基于仿真和实车平台对规划控制算法进行了验证。首先搭建了基于多机通讯的Prescan&Ros联合仿真平台,并设计园区典型工况对规划控制算法进行功能验证。然后基于哈弗H7实验车搭建实车测试平台,以工控机作为自动驾驶控制器,采用RTKGPS/IMU组合导航设备获取车辆的高精度定位及运动状态信息,通过CAN通讯实现与车辆的信息交互,并在校园内对规划控制算法进行了验证实验。仿真与实车实验结果均验证了本文算法的有效性和准确性。更多还原

【Abstract】 In the traditional "people-vehicle-road" transportation system,people are the weakest link and the main cause of traffic accidents.In the autonomous driving environment,because some or all of the driver links have been replaced,autonomous driving technology is considered to be a great solution to traffic safety issues.Among the application scenarios of autonomous driving technology,the park scene has attracted the attention of many enterprises and universities due to its low speed and simple working conditions.It is one of the first scenarios that autonomous driving can be implemented.The autonomous driving in the park has become a research hotspot in recent years.Relying on the key special project "Research and Demonstration Operation of Electric Autonomous Vehicles" by the Ministry of Science and Technology,this paper studies local planning and tracking control technology in the park scene,and performs algorithm verification based on simulation and real vehicle platforms.Aiming at the local planning problem,this paper first introduces the Frenet framework to decompose the local planning problem into two one-dimensional problems,which reduces the planning difficulty.Then considering the dynamic changes of the driving environment of the autonomous vehicle,this article analyzes the driving risk of the traffic environment,and models the driving environment based on the potential field method.By introducing adaptive adjustment factors,the risk influence range of dynamic obstacles can be adjusted automatically with the change of relative speed.Finally,the fifth-order polynomial curve is used to generate candidate trajectories.In this paper,the target states in the state space are reasonably sampled,and the lateral and longitudinal trajectory clusters are separately constructed and combined as a candidate set.This paper designs a multi-objective evaluation function that considers safety,comfort,and efficiency to evaluate candidate trajectories,select the optimal trajectory in this cycle.Repeat the above steps at the next moment,so as to realize the dynamic update of local planning.Aiming at the trajectory tracking problem,this paper adopts decoupling control strategies to build the vehicle lateral and longitudinal controllers respectively.Firstly,a vehicle lateral controller was built based on the model predictive control theory.The vehicle lateral dynamics model was used as a predictive model,and the constraints on control and output were designed.The MPC optimization problem was established and solved by transforming it into a quadratic programming problem.This method takes into account the accuracy and comfort of tracking.For longitudinal control,this paper designs a higher-level controller based on the optimal preview theory to achieve the desired acceleration decision.A lower-level controller is established through the vehicle’s longitudinal inverse dynamics model to achieve the conversion from acceleration to vehicle input,thereby achieving hierarchical longitudinal control Strategy.Finally,this paper validates the planning and control algorithm based on simulation and real vehicle platforms.First,a Prescan & Ros joint simulation platform based on multi-machine communication is built,and typical experimental conditions are designed to verify the function of the planning and control algorithm.Then build a real vehicle test platform based on the Haval H7 experimental vehicle,use the industrial computer as the automatic driving controller,and use RTK-GPS / IMU integrated navigation equipment to obtain the high-precision positioning and motion status information of the vehicle.The vehicle are controlled through CAN bus.The planning and control algorithm is verified in the campus.The simulation and real vehicle experimental results verify the effectiveness and accuracy of the algorithm in this paper.更多还原