Multi-sensors Fusion Localization Based On Millimeter Wave Radar

Achieving precise localization of the car in relatively harsh environments such as low light,rain,fog,dynamic changes,etc.is the basis for safe unmanned driving.Based on the advantages of different sensors,multi-sensor fusion localization is currently the consensus of scientific research and academia.Based on the millimeter-wave radar Doppler speed measurement capability,the vehicle's own speed and angular velocity information measured by the odometer are merged to calculate the global radial speed of the detected point,and the credibility of the radar detection is combined to realize the motion state of the object Detection.The detection of motion status is generally believed to greatly improve the stability of the positioning system.The experiments show that in a static environment,97% of the detection points have a global radial velocity of less than 0.5m/s when credibility threshold is 90%.When the credibility threshold is 90% and the radial velocity threshold is 0.5m/s,the accuracy of static point detection is 98.24%.When using millimeter-wave radar for particle filtering for localization,an odometer is required to provide the initial motion value.The odometer will almost fail in the event of a slip.The fusion of wheel speed odometer and IMU can alleviate the impact caused by skidding.The angular velocity measured by IMU and the angular velocity information calculated by the odometer are used to calculate the skid coefficient.The threshold is used to determine whether skidding is detected.Next,the acceleration information of the IMU is used to replace the speed information of the wheel speed odometer to update the position under the severe skidding.This fusion algorithm effectively improves the positioning accuracy.Through Simulink simulation experiments,it is shown that under severe skid conditions,even low-precision IMUs can still effectively prevent positioning collapse.The above odometer and IMU fusion algorithm can be used as a part of the millimeter wave radar particle filtering algorithm to provide the initial value of motion update.Particle filter positioning based on millimeter wave radar consists of two parts.The first part is the construction of the map.Using the high-precision odometer and millimeter-wave radar data,and based on the inverse sensor model of the millimeter-wave radar,the construction of the surrounding environment occupation grid map is completed,and the map is improved according to the space passed by the vehicle The second part is the particle filtering and localization part.It innovatively proposes a frame accumulation strategy and combines multiple radar layout methods to alleviate the problem of fewer millimeter wave radar points.An observation model is proposed.The particle filter algorithm is used to calculate the odometer.Data fusion with millimeter wave radar data effectively reduces the global cumulative error.Hyper-parameter experiments were carried out on the accumulated frames and resampling coefficients,which laid the foundation for further research.In order to carry out the above research,through Simulink simulation and real vehicle test,the program and algorithm were verified to fully explore the role of millimeter-wave radar in the field of sensor positioning and the ability to integrate with other sensors such as odometers to reduce the cumulative error.