Research On The Learning Of Longitudinal Dynamics Characteristic And The Control For Unmanned Driving Robot Vehicle

The driving robot is adapted to different types of vehicle due to its flexible structure.The horizontal and vertical coordinated control of unmanned driving robot vehicle(UDRV)is studied by our research group.On the basis of this,to better operate different types of vehicle under the premise of understanding the dynamics characteristics for the operated vehicle,the learning methods of longitudinal dynamics characteristic and the longitudinal dynamic control are studied.Firstly,the performance indexes are given,and the structural characteristic of shift mechanical manipulator and driving mechanical legs are analyzed.On the basis of this,kinematics and dynamics models of each mechanical manipulator are established.Then,the offline and online learning methods of longitudinal dynamics characteristic for UDRV are introduced.The driving torque and transmitted torque for clutch engagement in the process of shift are estimated by kalman model.And the braking torque is estimated by extended autoregressive model,which are identified online by fuzzy variable forgetting factor recursive least square algorithm(FFRLS).Then,two kinds of adaptive tracking control methods to multi-modal speed for UDRV are introduced.One is fuzzy supervised control,the other is the combination of dynamic fuzzy neural network(DFNN)and proportional-integral(PI),which are model-dependent and model-free,respectively.The corresponding control laws are given and their stabilities are proved theoretically.Finally,in view of the fact that the dynamics characteristic of UDRV changes in a long driving,the dynamic control methods for UDRV to improve the driving quality are proposed based on learning methods of the longitudinal dynamics characteristic.And the control of engagement for clutch mechanical leg has influence on the shifting performance.A finite-time linear quadratic regulator(LQR)of the engagement control for clutch mechanical leg is designed considering the friction work and shift jerk.A neural network adaptive robust controller of driving mechanical legs is designed to track quadratic optimal transmitted torque.On the basis of this,the effective of the longitudinal dynamic control methods for UDRV is proved by the experiments.