Study On Autonomous Planning Of Unmanned Vehicle Speed Based On Intelligent Learning On Complicated Pavement

Unmanned Ground Vehicles(UGVs)are the hot spot in the field of artificial intelligence,which have attracted extensive attention from both foreign and domestic research institutions and automobile manufacturers.Current automatic driving technologies are mainly focused on urban road environments.Generally,the characteristics of ground environments are simple and flat.Compared with the urban road environment,the outdoor’ is usually more complicated and bumpy(such as gravel road,grassland,dirt road,etc.).Therefore,considering the factors of driving stability and safety,different driving speeds need to be planned according to different size of bumps.As a consequence,self-piloting technologies of higher requirements are needed to put forward.Depending on which sensor is relied upon,common methods of bump identification generally based on internal sensors or external sensors.External sensors(such as laser,image data)are able to take advantage of abundant external environment information.However,the results of the analysis are easily affected by the external environment.Compared with external sensor,the internal sensor(such as inertial navigation data)has higher detection frequency and better stability.Therefore,how to take advantage of the internal sensor information to distinguish the degree of bumpiness of different terrain and make a reasonable speed grade strategy is the main content of this paper.Specifically,it can be divided into the following aspects.First,based on the North Star(razor-800 all-terrain SUV),the UGV was built.In the hardware part,the modification and debugging of the bottom actuator such as brake,throttle and gear was achieved.According to the different sensor’s function and the characteristic,the corresponding installment and the wiring was carried on.In addition,according to the size and power supply of different devices,the structure of the power supply box and main control box was designed.In the software part,the environment-aware,modeling,planning and underlying control modules were development.Data communication among multiple modules was built by the way of UDP.The real-time analysis of the sensor information number and real-time generation of the bottom control signal was realized.Second,Estimate the degree of bumpy road was achieved through analyzing the intrinsic characteristics of the acceleration signal collected by the inertial navigation device.Specifically,the method(Wavelet decomposition noise reduction)processed the collected acceleration signal,and through the singular value analysis,extract the main features of the signal after noise reduction.On this basis,taking the fastness and accuracy of the method into consideration,an Extreme Learning Machine(ELM)was used as a classifier.This method distinguishes the turbulence of different ground environments and overcome the problems of traditional methods such as support vector machine(SVM)sensitive to nuclear parameters and high computational complexity and realize the autonomous speed planning of the unmanned platform based on the given driving speed level.