Research On Automatic Driving AEB Control Strategy And Test Evaluation Considering Braking Intention Of Front Vehicle

As one of the active safety technologies for autonomous driving,the AEB system can effectively avoid the occurrence of rear end collisions.However,due to the limited perception ability and poor adaptability of control strategies of the AEB system,it is easy to cause problems such as delayed vehicle braking and high false alarm rates.Based on this,this article conducts research on the subjective braking intention of the driver in front of the vehicle,establishes a braking intention recognition model based on double-layer HMM,proposes an AEB control strategy considering the braking intention of the front vehicle,and to verify the effectiveness of the proposed control strategy in the braking scenario of the front vehicle,a joint simulation testing platform is built,and a comprehensive evaluation method for collision avoidance performance of the AEB system based on Analytic Hierarchy Process is proposed.The specific research work of this article is as follows:(1)Build a PreScan/Simulink/Driving Simulator joint simulation sampling platform to collect brake pedal pressure,brake pedal pressure change rate,and vehicle speed data of real drivers driving with slight/normal/emergency braking intentions.Use K-means clustering method to classify the braking intention of the front vehicle and determine the true value of the driver’s braking intention during driving.Using a sliding time window with multiple parameter combinations to extract fixed step historical data sequences,establish a training dataset for braking intention recognition model parameters.(2)Considering the impact of front vehicle braking intention on the main vehicle AEB control strategy,a front vehicle braking intention recognition method based on a double-layer hidden Markov model is proposed.Firstly,classify and quantify various types of data based on the distribution of the original sampling data.Secondly,establish a braking behavior recognition HMM and braking intention recognition HMM,and train the model based on the Baum Welch algorithm.Finally,using accuracy,recall,F1 score,and accuracy as model evaluation indicators,the optimal intention recognition model is determined by comparing the recognition results of the driver’s braking intention using the double-layer HMM model established with different parameter combinations.(3)Considering the impact of communication delay and vehicle braking delay on vehicle collision avoidance process in the context of vehicle networking,an AEB control strategy based on the intention of front vehicle braking is proposed.Firstly,study the specific effects of critical safety threshold,communication delay,and brake pressure establishment time on the collision avoidance performance of the AEB system during operation.Secondly,analyze the motion status of the front and rear vehicles during the active collision avoidance process.Finally,design an AEB control strategy that can calculate the critical safety distance threshold under current operating conditions and conduct hazard assessment and decision control based on different braking intentions of the front vehicle.(4)A comprehensive evaluation method based on Analytic Hierarchy Process(AHP)is proposed to address the issues of single and one-sided evaluation indicators in existing AEB evaluation methods.Taking vehicle safety and comfort as the evaluation objectives,seven specific evaluation indicators are determined,and the AHP method is used to calculate the weight of the indicators and conduct consistency testing.Based on the PreScan/Simulink joint simulation platform,simulation testing was conducted,referencing testing standards such as ENCAP and CNCAP.Test scenarios covering low,medium,high speed,and different braking intentions were set up,and the test results of four typical AEB models,Mazda,Honda,Berkeley,and TTC,were compared to verify the effectiveness of this model in improving vehicle safety.The specific experimental results are as follows: 1)In the front car brake intention recognition experiment,the data model extracted using sliding time windows with combined parameters of 0.05 s,0.2s,and 0.8s has the best training results,and the recognition accuracy for the three types of brake intentions can reach 97.38%,proving the effectiveness of the double-layer HMM model for brake intention recognition.2)In the simulation testing experiment of the AEB model,based on the comprehensive evaluation method proposed in this article,the AEB model considering the braking intention of the front vehicle obtained 81.85 points,which is much higher than other AEB models,proving that the model can avoid rear end accidents under different braking intentions of the front vehicle,while also avoiding driving discomfort caused by premature braking of the vehicle.