Research On Collision Avoidance Of Intelligent Vehicles Based On States Estimation

Intelligent vehicle(IV) is a major part of intelligent transport system. It is related to computer science, advanced sensor technology, information fusion, communication, artificial intelligence, automatic control and so on. Drivers are liberated from heavy driving work, the ultimate goal is driverless vehicle. It is benefit for enhancing the traffic efficiency, decreasing congestion, energy consumption, reducing pollutant emissions. IV is the new hot research field of world vehicle engineering and new energy of automotive industry. Therefore, there is a significant academic and practical value of its deep study.With the development of the intelligent vehicle, a lot of key problems need to be solved. Especially, collision avoidance and driving safety in the bad weather and slippery roads. Collision avoidance of intelligent vehicle is on the basis of obtaining real-time vehicle state and tire-road friction coefficient. The vehicle state is measured by on-board sensors as a rule, but due to the vehicle is always driving in complex environment, sensors’ measurement accuracy and cost, the key states of the vehicle are hard to directly measured, which limits the development of collision avoidance key technology.This thesis focuses on the vehicle state and friction coefficient estimation, collision avoidance, local motion planning and ABS control method of collision avoidance, which based on the 6-DOF vehicle HIL simulation platform, and intelligent vehicle of Beijing University of Technology BJUT-IV. The major content is summarized as follows:(1) The vehicle dynamic model and vehicle kinematic model are developed. The dynamic models includes four-wheel eight degree of freedom vehicle dynamic model and bicycle model. Pacejka tire model, Dugoff tire model and LuGre dynamic tire model are developed for analysis tire dynamics. The kinematic vehicle model is built based on the vehicle geometry. The CarSim software is used to evaluate the model accuracy.(2) Due to the vehicle state and friction coefficient are hard to measure directly, An overall estimation method is proposed to vehicle state estimation. The four degree of freedom vehicle model combined with Dugoff tire model and LuGre dynamic tire model are used as the models of estimation, respectively. The dual-UKF is explored for estimate the vehicle state and friction coefficient. Compared between EKF and UKF estimation result in the strong nonlinear simulation scenarios. The friction forces estimation results are compared between using LuGre model and using Dugoff tire model, respectively.(3) To deal with tire working in the small slip ratio zone, the vehicle and tire estimation method is hard to measure the friction coefficient, and using the intelligent tire to estimate the friction coefficient impact by the tire speed. The tangential acceleration signal power spectrum method is proposed to classify different road surfaces. The power amplitude and power rate are chose as the feature of classification. The support vector machine(SVM) is used to regress the friction coefficient. Then, the multi-information decision fusion is addressed to identify the more accuracy friction coefficient, which combined the vehicle and tire model estimation method and intelligent tire method.(4) For collision avoidance system always ignore the road surfaces condition and braking is the only way to mitigate the collision. The adaptive warning and braking method is developed for collision avoidance system in the single lane situation. When the environment around vehicle is safe, the local collision avoidance trajectory is developed by ripple tentacle method. The time varying model predictive control method with the kinematic model and constrain is used to track the local trajectory.(5) In order to adapt the different surfaces and tire pressure distribution for ABS controller, the multiple models backstepping adaptive control method is proposed for ABS control. The high friction fixed model, middle friction fixed model, low friction fixed model and the adaptive ABS model are composed of the multiple models. The different simulation scenarios are used to make ABS controller validation, the stability of controller is proved by the Lyapunov stability theorem.