Research On Autonomous Emergency Braking Strategy Based On Machine Vision Road Surface Recognition

The autonomous emergency braking system is a typical intelligent driving assistance system,which can warn the driver to brake in advance,and automatically brake when necessary,to prevent collisions or minimize collision injuries.However,the current autonomous emergency braking system design mode is single,ignoring the influence of road surface,unable to adapt to the differences of different road surfaces,and reducing the driver’s acceptance and satisfaction of autonomous emergency braking system.As the necessary information in the driving process,the state of the road surface plays an important role in guiding the formulation of braking strategies.Therefore,in-depth research on road surface recognition algorithms and targeted design of autonomous emergency braking strategies with road surface recognition are the research directions of autonomous emergency braking systems and are also necessary for large-scale application and expansion of autonomous emergency braking systems.Relying on the Natural Science Foundation of Jilin Province(No.20210101057JC),this paper conducts research on autonomous emergency braking strategies based on machine vision road surface recognition.Build a real vehicle driving data collection platform,collect driving scene pictures including road surfaces,build a driving scene dataset,and use semantic segmentation algorithms to identify driving scenes including road surfaces;establish a corresponding relationship model between road types and adhesion coefficients,and design a road surface decision algorithm based on semantic segmentation results;oriented to safety and comfort requirements,an automatic emergency braking strategy is designed;the driving scene semantic segmentation algorithm and the road surface decision-making algorithm are tested using the collected real picture data.Finally,a Python-Matlab/Simulink-Car Sim co-simulation platform is built to verify the automatic emergency braking strategy based on road surface recognition.The main research content of this paper includes the following four parts:(1)Driving scene data collection and driving scene semantic segmentation algorithmFirst,based on software and hardware such as cameras and host computers,a real vehicle driving data collection platform is built to collect driving scene data including structured and unstructured road surfaces.The data were preliminarily screened and processed according to the general image screening principles.Define semantic segmentation model categories based on target requirements and collected data types.Finally,a driving scene semantic segmentation algorithm is designed.(2)Road surface decision algorithm based on semantic segmentation resultsConstruct the road surface type and adhesion coefficient mapping relationship table to obtain the road surface type and adhesion coefficient mapping relationship.Using the results of semantic segmentation,establish a region of interest(Region of Interest,ROI),in which the road surface type is counted,and the road surface set in the current area is obtained.Using the established correspondence between road surface types and adhesion coefficients,the adhesion coefficient set corresponding to the current road surface is obtained.Finally,the road surface screening rules are designed to obtain the road surface adhesion coefficient in this surface.The offline simulation verification of the pavement state decision-making algorithm is carried out in typical working conditions,and the results show that the algorithm is effective.(3)Autonomous emergency braking strategy based on road surface recognitionUsing the road surface decision algorithm,the current road surface adhesion coefficient is obtained.According to the adhesion coefficient of the road surface,an automatic emergency braking strategy for graded braking is generated.The PID control algorithm is used to complete the conversion of the braking deceleration to the pressure of the master cylinder to complete the braking.The algorithm improves the safety of the autonomous emergency braking system under different road surfaces and improves the utilization rate of the traffic environment.(4)Test verification and analysisBuild a test and verification platform and use real collected data to verify the semantic segmentation algorithm and the road surface decision algorithm.According to the European New Car Assessment Program(Euro-NCAP)standard,the automatic emergency braking strategy is verified under different working surfaces.Build the Transmission Control Protocol(TCP)to realize the communication between the Python platform and Simulink.At the same time,the communication between Simulink and Carsim platform is completed.Finally,based on the Python-Matlab/Simulink-Car Sim platform,the autonomous emergency braking strategy based on road surface recognition is verified.The results prove the safety,reliability and feasibility of the automatic emergency braking strategy based on road surface recognition.