Nonlinear Estimation Of Vehicle State And Tire-road Adhesion Coefficient

The driving safety of vehicles has become a global social problem. Active safetycontrol systems can improve the handling and stability of vehicle and then avoid trafficaccidents. Many of these systems have in common that the control action depends oninformation about longitudinal and lateral velocities, friction forces and tire-road adhesioncoefficient. However, these variables cannot be measured directly in modern cars. As aconsequence, the estimation of them has attracted the attention of many researchers overthe past few years due to its important theoretical and practical significance.In this dissertation, the problems of nonlinear estimation of vehicle state (includinglongitudinal and lateral velocities, and yaw rate), friction forces and tire-road adhesioncoefficient are discussed. For more systematization and completeness, a nonlinear vehicledynamic model is developed firstly. Then the estimation of vehicle state is mainly carriedout in three cases: on ?at roads with knowledge of tire-road adhesion coefficient, onnon-?at roads without knowledge of tire-road adhesion coefficient, and on non-?at roadswith knowledge of tire-road adhesion coefficient. In each case, the nonlinear estimationstrategy is respectively proposed with stability guarantees and performance validation.Finally, an effective nonlinear estimation method is presented for simultaneous estimationof longitudinal and lateral velocities and tire-road adhesion coefficient. Moreover, all ofthe methods proposed in this dissertation can be used to estimate friction forces.As the basis of developing estimation methods, a nonlinear vehicle dynamic modelis established in Chapter 2. In this model, not only the longitudinal, lateral and yawmotions of the vehicle body, but also the effect of static and dynamic weight distribution,coupling between longitudinal and lateral friction forces and force saturation are takeninto account. It is a comparatively comprehensive model and is convenient for analysisand calculation. The accuracy and effectiveness of the model are validated by simulationanalysis of the tire model and experimental tests of the vehicle dynamic model.According to the nonlinear dynamic characteristics of vehicle model, Chapter 3presents a nonlinear observer which can solve the problem of vehicle state estimationon ?at roads with knowledge of tire-road adhesion coefficient. A sufficient condition isderived to guarantee the stability of the observer based on Lyapunov stability theory, and the robustness of the observer with respect to additive disturbances is analyzed with thehelp of input-to-state stability (ISS) theory. Then, taking a test vehicle as example, theselection of the observer gains is discussed. Furthermore, the performance of the observeris compared with that of existing methods and evaluated analytically and experimentallyunder a variety of maneuvers and road conditions.To solve the problem of vehicle state estimation on non-flat roads without knowledgeof tire-road adhesion coefficient, a nonlinear estimation method for vehicle state withroad angle and tire-road adhesion coefficient adaptation is proposed in Chapter 4. Byestimating longitudinal friction forces and treating the acceleration measurement biasescaused by road grade and bank angles as unknown inputs, a vehicle state observer isdesigned based on sliding mode unknown input estimation technique. The observer canestimate vehicle state and reconstruct road angle without knowledge of tire-road adhesioncoefficient. In addition, the stability of the proposed method can be guaranteed in theory.If the tire-road adhesion coefficient is known, the proposed method can be simplified asa nonlinear estimation method for vehicle state with road angle adaptation. Finally, theperformance of the proposed methods is validated by simulation and experimental testsand compared with that of the method proposed in Chapter 3.Aiming at obtaining the tire-road adhesion coefficient, which has a close relationshipwith the nonlinear dynamic characteristics of the longitudinal and lateral vehicle motions,a moving horizon estimation (MHE) strategy for longitudinal and lateral velocities andtire-road adhesion coefficient with stability guarantees is presented in Chapter 5. Theproposed MHE strategy can take system disturbances caused by pavement roughness intoaccount and make use of additional knowledge to achieve improvements in the estimationperformance. Then the proposed MHE strategy is validated under a variety of maneuversand road conditions and compared with an extended Kalman filter strategy implementedfor the same purpose. The simulation and experimental results show that the proposedMHE strategy is effective for estimation of vehicle state and tire-road adhesion coefficient.