Research On Vehicle Localization Under NLOS Environment Based On Self-built Acoustic Database

Self-driving vehicles will improve road safety and offer new mobility options to millions of people,and one of the core technology that achieves self-driving is perception.However,most of the current perception systems equipped on vehicles can only detect objects within the line of sight,which makes it difficult for vehicles to avoid traffic accidents under non-line of sight conditions.This paper proposes a non-line-of-sight vehicle noise location method based on sound scattering for vehicles driving in blind areas.This method can locate vehicles with noise characteristics in a self-built database,which could protect from the accident.The main work of this paper includes the following aspects:First of all,according to the theory of vehicle vibration and acoustics,the main source of vehicle noise is determined,and the spectral characteristics of the noise are analyzed,and the vehicle noise identification is realized based on the self-built database.Based on highway engineering technical standards,combined with the characteristics of vehicle noise in the database,the driving state of the vehicle is indirectly determined,which is provided for the location of the vehicle’s non-line-of-sight sound source.Secondly,through the statistics of the propagation characteristics of the acoustic signal under different conditions,the gain and delay information of the signal are obtained.By extracting relative feature values and selecting the optimal classifier,the recognition of non-line-of-sight acoustic signals can be realized,and the interference and influence of other types of acoustic signals on positioning can be reduced.Finally,based on the characteristics of sound wave diffraction and Doppler frequency shift,triangulation and geometric relations are used to locate the corner and the position of the car,and the multi-sensor fusion algorithm is used to correct the positioning error accordingly,so as to achieve non-line-of-sight single sound source localization and tracking of vehicles traveling in a straight line at a constant speed.This approach can realize vehicle location positioning in NLOS scenarios with four microphones and increase driver safety.The effectiveness of the proposed method is validated by simulation based on the finite element method.