Research On The Longitudinal Dynamics Robust Control Method Of Vehicles Based On Multi-sensor Fusion

In recent years,the self-driving technology has been developing rapidly and has become a hot topic in the automotive field.Self-driving vehicles on the market already have the ability to achieve self-driving operation in some typical scenarios,however,facing some unexpected situations,it is still not possible to fully guarantee the safety of the passengers.The main reason is that the perception system and control system of the vehicle are vulnerable to the interference of environmental noise and system noise,resulting in a large error in vehicle perception and control.To resolve the above problems,this paper researches on the robust control method of vehicle longitudinal dynamics based on multi-sensor fusion,and the main research is as follows:(1)Estimation of vehicle dynamics states.Considering that it is difficult to measure the vehicle state in practice,such as the vehicle velocity,tire-road friction coefficient,tire stiffness,etc.,three modular observers are designed to estimate the driving state of the vehicle in order to improve the accuracy and robustness of the observation.Firstly,for the wheel rotation speed sensor in practice is susceptible to environmental noise interference,a observer of wheel rotation speed is designed based on the wheel dynamics model,and the stability of this observer is demonstrated based on Lyapunov theory;Then,considering the interference of additive noise in the environment,a nonlinear state observer is designed for the vehicle’s lateral velocity,longitudinal velocity,yaw rate and tire-road friction coefficient based on the vehicle seven-degree-of-freedom dynamics model and the Dugoff tire model.Moreover,the gain range of the observer is solved based on the Lyapunov stability,and the convergence and robustness of the designed observer are demonstrated;Finally,based on the Dugoff tire model,the longitudinal stiffness and cornering stiffness of the tire were estimated using the Recursive Least Squares(RLS)algorithm,and a forgetting factor was introduced to adjust the convergence rate of the observations.(2)Multi-sensor data fusion.Considering the dynamic uncertainty of the environment and the random interference of the sensor system,a novel multi-sensor data fusion method is proposed to improve the redundancy,robustness and accuracy of the perception system.Firstly,based on the Hungarian algorithm to associate the obstacles sensed by multiple sensors and reduce the fusion error caused by the target matching failure of multiple sensors;Then,considering the disturbances such as system modelling uncertainty and environmental noise uncertainty,a mixed H∞/AKF data fusion method is proposed by integrating the advantages of H∞ filtering and Adaptive Kalman Filter(AKF),together with the introduction of an attenuation factor in AKF to avoid system divergence caused by the accumulation of errors due to environmental noise.Finally,considering the lateral offset due to road curvature,the selection logic of Closest In-Path Vehicle(CIPV)of this lane is designed and the multi-sensor fusion information of CIPV is utilized for vehicle longitudinal dynamics control.(3)Longitudinal dynamics control of the vehicle.For the uncertainty problem of vehicle system,considering the safety,comfort and other constraints of intelligent vehicles,the robust optimal control algorithm of vehicle longitudinal dynamics is proposed,and the robustness design criterion of vehicle longitudinal dynamics is established based on H∞control theory to realize the robust control of vehicles.Firstly,the curvature of the road is estimated based on the Kalman Filter(KF)algorithm,which fuses sensor-perceived lane information and vehicle dynamics information;Then,the following time-distance and safe velocity are designed based on the fused road curvature to realize the adaptive control of the vehicle in the following scenario,Finally,considering the perturbations of chassis system uncertainty and system modeling uncertainty,an H∞-based closed-loop controller for vehicle longitudinal dynamics is designed using Lyapunov stability theory.(4)Hardware-in-the-loop verification.A variety of different hardware-in-the-loop test scenarios were established for the validation of the proposed method in this paper.The results show that the proposed robust control method based on multi-sensor fusion for vehicle longitudinal dynamics can meet the requirements of vehicle comfort,safety,robustness and real-time,and can be applied to real-time vehicle controllers and intelligent driving scenarios.