| Increasing the automatic emergency braking system configuration rate of passenger vehicles has an important practical significance for improving the level of road traffic safety and protecting the lives and properties of drivers and passengers.This paper focuses on the development and verification of AEBS and its control strategy in electric vehicles,committing to designing a set of advanced driving assistance system with high integration,strong robustness and higher control accuracy to adapt to the intelligent and electrified vehicles.Based on this,the research contents of this paper are as follows:Aiming at the shortcomings of the VGG16 feature extraction network,such as long training time,and a large number of features etc,the Faster-RCNN algorithm is improved by combining a small calculation and easy-to-optimize deep residual network(Res Net50)with feature pyramid network(FPN).The training and validation of the improved Faster-RCNN algorithm model are completed by using INRIA dataset,MIT dataset and nu Scenes dataset.Relying on the test samples collected from city roads,tunnels and inner ring highways to verify the detection accuracy and generalization ability of the improved Faster-RCNN algorithm model.The experimental results show that the improved Faster-RCNN algorithm has strong robustness and the detection speed on GPU can reach 2-4 FPS,which can achieve the detection of target objects in complex scenes such as occlusion and curves.To solve the problem of inconsistent sensor data in spatial dimensions caused by the different installation positions and sampling frequencies of millimeter wave radar and camera on the vehicle,the spatial conversion model between radar coordinate system-camera coordinate system-image coordinate system-pixel coordinate system is established by deriving the spatial conversion equations of millimeter wave radar coordinate system and image coordinate system.Then the calibration of Logitech C670 i camera is completed combining Zhengyou Zhang’s camera calibration method.Moreover,the fusion of radar and camera data in spatial dimensions is realized,and the accuracy of fusion model is verified via the experiments.A layered warning model is built integrating safe driving level,collision warning level and emergency braking level,determined the thresholds of time to collision and driving safety level under different vehicle speeds.A layered control algorithm model with upper single hidden layer neural network control and lower PID control is proposed,and using the experienced drivers’ braking data to train the neural network model.Aiming at the problem that the PID parameters adjustment process is time-consuming and overly relies on the engineering experience,according to the random walk theory,improved the learning factors of particle swarm optimization algorithm and used it to tune PID parameters automatically,which improves the control accuracy and self-adjustment ability of this system.What’s more,this paper integrates the vehicle driving relationship and powertrain structure to derive the expressions of the inverse dynamics model of the vehicle,and establishes the conversion model of the desired deceleration to the motor torque and brake master cylinder pressure,the logic switching model of braking and acceleration conditions,electric vehicle dynamics model,millimeter wave radar model and camera model etc.The software modeling of automatic emergency braking system and its control strategy is completed.According to the vehicle active safety technology test scenarios within C-NCAP(China-New Car Assessment Program,2021 version),the simulation scenarios are built using Car Sim software,and the feasibility and effectiveness of the overall framework of the electric vehicle automatic emergency braking system and its control algorithm designed in this paper are verified by combining MATLAB/Simulink software with Car Sim software and the overall evaluation of the functional safety in the electric vehicle automatic emergency braking system is carried out. |
PDF Full Text Request