摘 要

汽车自动紧急制动系统（AEB）是汽车辅助驾驶系统中的一种，其作用时能够在驾驶员来不及反应的紧急情况下自动采取紧急制动措施以避免碰撞或减小碰撞损失，AEB系统的装备对减少交通碰撞安全事故，提高驾驶安全有着重要意义。

无人驾驶是目前汽车发展的趋势，AEB作为重要的主动安全技术，是无人车不可或缺的一部分。本文主要基于Apollo D-kit无人线控底盘研究了CCR工况，复杂路面工况，直道、弯道、出弯道入直道多目标工况下的AEB控制策略，其中包括路面情况识别、有效车辆目标筛选、以及车辆状态判断，并基于PID理论控制制动压力输出，基于模糊控制实现制动能量回馈，搭建了MATLAB和CarSim的联合仿真模型，以及进行单片机小车测试，通过仿真和试验结合的方式验证了AEB控制策略的可行性，得出以下结论。

（1）基于最小二乘法对路面附着系数进行识别，实现对附着系数的跟随。滑移率较小时计算轮胎的纵滑刚度来匹配路面；滑移率较大时利用魔术轮胎公式，确定公式中三个常量的值，根据利用附着系数和滑移率对路面进行判断。仿真验证了在冰雪，泥土，沥青以及对接路面上AEB算法能够对路面附着系数进行识别并将信息传达给状态判断控制器中。

（2）基于碰撞时间倒数（iTTC）和危险系数结合的方法对车辆运动状态进行判断，能够有效的达到避撞的效果。危险系数由安全距离计算得到，能够有效实现体现路面对车辆运动状态判断的影响，iTTC能够有效划分碰撞等级。仿真验证了两者结合能够有效对车辆状态进行判断，并执行分级制动和分级预警措施，满足安全性的同时兼顾舒适性。

（3）基于横向距离判断方法，能够实现在直道、弯道、出弯多车道多场景下的危险目标判断及筛选。直道场景下采用判断目标车辆与自车的横向距离大小进行判断；弯道场景采用将自车位置补偿到与目标车同一条半径上进行两车弯道半径的比较；出弯场景则是采用将车辆状态离散化的方法分析车辆所在的车道半径。仿真验证了AEB算法在多目标工况下对于非自车道的车辆，自车道的车辆以及突然换道的车辆，系统能够有效的识别。

（4）基于模糊控制搭建的Apollo D-kit线控底盘模型能够实现期望速度跟随以及在紧急制动过程中实现能量回馈。线控底盘模型包括电机、电池和底盘控制模型，根据电池SOC值、车速、制动减速器三个因素对制动力分配系数进行控制。仿真验证了AEB算法在CCR工况下实现制动能量的回馈，以及对电机开度进行控制，实现对目标车速的跟随。

（5）采用基于STC89C52芯片控制和蓝牙模块无线信息传输的智能小车进行测试，能够实现避撞功能。运用Simulink自动生成C代码模块生成状态判断模块的代码，运用蓝牙模块将信息传输到PC端中，测试表明在遇到障碍物时小车能够发出报警信号同时切断动力有效的避免碰撞，验证了AEB算法的可行性。

关键词：自动紧急制动；AEB；目标识别；路面识别；安全距离

Abstract

Automobile emergency braking system (AEB) is a kind of automobile assisted driving system, which can automatically take emergency braking measures to avoid or reduce collision losses in emergency situations where the driver has no time to respond. Equipment with AEB system is of great significance to reduce traffic collision safety accidents and improve driving safety.

Self-driving is the current trend of automobile development. As an important active safety technology, AEB is an indispensable part of unmanned vehicles. This paper mainly studies the AEB control strategy based on Apollo D-kit unmanned remote control chassis under CCR conditions, complex pavement conditions, straight road, curved road, out of curve road under multi-target working conditions, including road surface identification, effective Vehicle target screening and vehicle state judgment, based on PID theory to control the brake pressure output, based on fuzzy control to achieve braking energy feedback, built a MATLAB and CarSim joint simulation model, and single-chip car test, through simulation and test combined the method verifies the feasibility of the AEB control strategy and draws the following conclusions.

(1) Identify the pavement adhesion coefficient based on the least square method to achieve the following of the adhesion coefficient. When the slip rate is small, the longitudinal slip rigidity of the tire is calculated to match the road surface; when the slip rate is large, the magic tire formula is used to determine the three constant values in the formula, and the road surface is judged based on the use of the adhesion coefficient and the slip rate. The simulation verifies that the AEB algorithm can recognize the pavement adhesion coefficient and transmit the information to the state judgment controller on ice, snow, soil, asphalt and docking pavement.

(2) Based on the method of combining the inverse collision time (iTTC) and the hazard coefficient, the vehicle motion state can be judged, which can effectively achieve the effect of collision avoidance. The hazard coefficient is calculated from the safety distance, which can effectively reflect the influence of the road surface on the judgment of the vehicle's motion state. ITTC can effectively divide the collision level. The simulation verifies that the combination of the two can effectively judge the state of the vehicle, and implement hierarchical braking and hierarchical early warning measures to meet safety while taking into account comfort.

(3) Based on the lateral distance judgment method, it can realize the judgment and screening of dangerous targets in straight roads, curved roads, and multi-lane multi-scene scenes. In the straight road scenario, the horizontal distance between the target vehicle and the own vehicle is used to judge; in the curve scene, the position of the vehicle is compensated to the same radius as the target vehicle to compare the two vehicle bend radius; The vehicle state discretization method analyzes the lane radius where the vehicle is located. The simulation verifies that the AEB algorithm can effectively identify the vehicles that are not in the lane, vehicles in the lane, and vehicles that suddenly change lanes under multi-target conditions.

(4) The Apollo D-kit wire-controlled chassis model based on fuzzy control can achieve the desired speed following and energy feedback during the emergency braking process. The wire-controlled chassis model includes a motor, battery, and chassis control model, and controls the braking force distribution coefficient according to three factors: battery SOC value, vehicle speed, and brake reducer. The simulation verifies that the AEB algorithm realizes the braking energy feedback under the CCR operating condition, and controls the motor opening degree, so as to follow the target speed.

(5) A samll car based on STC89C52 chip control and Bluetooth module wireless information transmission is used for testing, which can realize the collision avoidance function. Using Simulink to automatically generate the C code module to generate the code of the state judgment module, and using the Bluetooth module to transmit the information to the PC, the test shows that the car can send an alarm signal while cutting off the obstacle and cut off the power to effectively avoid collision, and verified the AEB algorithm feasibility.

Key Words：Automatic emergency braking; AEB; target recognition; road recognition; safe distance

目 录