Research On Emergency Braking Intention Recognition Based On Driver's Lower Limb EMG Signal

With the popularization of automobile,the transportation system consisting of people-car-road is constantly going in the direction of mega-complexity and complexity,and the road traffic safety problem becomes more severe and urgent.The statistical results show that more than 90% of road traffic accidents at home and abroad are inseparable from human factors.Emergency braking,as a typical driving behavior,usually involves dangerous traffic conditions,directly related to the safety of road traffic.Therefore,the driver's emergency braking intention research has always been one of the important research contents in the field of active safety of vehicles.In the past,the research on driver's braking intention mainly uses vehicle running state data and driving operation information as the identification parameters,and recognizes from the end of intended action chain,so there is a certain room for improvement in the recognition efficiency.In response to this problem,this study introduces the surface electromyography(sEMG)of driver's lower limb muscles as a recognition parameter into the identification model of driver's emergency braking intention.The aim is that the driver's emergency braking intention can be recognized before a driver steps on the brake pedal,and the brake assist system can be triggered in time to ensure driving safety.Firstly,the acquisition,preprocessing and feature extraction methods of sEMG were compared and analyzed.For emergency braking action,conventional braking and acceleration shifting action which are very similar to the characteristics of emergency braking,a feature vector is the composition of some features with higher separability,which are selected from multiple dimensions,such as time domain,frequency domain,time-frequency domain and AR parameter.And the vector is used as an input to the emergency braking intent recognition model.Then,in view of the advantages of high efficiency and good classification performance of support vector machine(SVM),the SVM multi-classification algorithm based on directed acyclic graph is used to establish the driver emergency braking intention recognition model.Aiming at the parameter selection problem in the model,a memetic algorithm based on particle swarm optimization and hill-climbing search is designed.The memetic algorithm is used to optimize the model parameters,and can balance the global evolution and local search ability,and ensure the model has a good performance.Finally,for safety reasons,a driving simulator was used to conduct simulated driving experiments.The experimental data was used to verify the validity of the established emergency braking intention recognition model.The results show that the final recognition rate of emergency braking intention of same individual is up to 92.3%,and that of different individuals can reach 85.6%.Moreover,the system can detect emergency braking intention 220 ms earlier than operating brake pedal.At 100 km/h driving speed,this amounts to reducing the braking distance by 6.1m.