Research On Adaptive And Learning Control Of Wheeled Mobile Robot

As a typical representative of the intelligent robot field,wheeled mobile robots have the advantages of simple mechanical structure,great movement flexibility,good operating performance,and high energy utilization rate.They are more active in the fields of space exploration,industrial manufacturing,living services,agricultural production and so on.The characteristics of non-linearity,under-driving,and non-holonomic constraints of wheeled mobile robot bring difficulties and challenges to its tracking control.In complex work scenarios,mobile robots are not only restricted by various constraints but also affected by external interference and the uncertainty of the parameters in the system,which puts higher requirements on the controller design.In addition,for repetitive work tasks,learning and acquiring knowledge from experience and using it to design learning controllers can improve control performance and reduce energy consumption.When performing complex control tasks,multirobot cooperation is equipped with higher work efficiency.In this paper,based on neural network and adaptive control theory,the trajectory tracking control method of wheeled mobile robot is studied by combining Lyapunov stability theory,chain transformation,backstepping method,obstacle Lyapunov function,deterministic learning theory,and consistency control theory.The main research contents of the paper are as follows:(1)Considering the actuator dynamics,an adaptive neural network trajectory tracking controller is designed using the backstepping method for the uncertain wheeled mobile robot with external interference.The designed controller can achieve the tracking for the desired trajectory and realize local accurate approximation(learning)of the unknown closed-loop system dynamic in the stable tracking process.Using the learned knowledge to construct a static learning controller can improve the control performance of the system and reduce energy consumption.(2)Considering input saturation and actuator dynamics,for the uncertain wheeled mobile robots with external disturbance,by introducing a suitable dynamic system to tackle the impact of input saturation,an adaptive neural network trajectory tracking controller is designed.While ensuring that the closed-loop system achieves the tracking of the desired trajectory,the control input of the system remains within the constraints.And it is verified that the introduction of dynamic system does not affect the learning of unknown closed-loop system dynamics in the control process.The learning controller constructed subsequently proved to show its efficiency and superiority in the same control task.(3)Considering state constraints and actuator dynamics,for the uncertain wheeled mobile robot with external disturbance,a virtual control law and an adaptive neural network trajectory tracking controller are designed based on the chain model.While the closed-loop system achieving stable tracking,the system states are also within the constraints.In the stable adaptive control process,the local accurate approximation(learning)of the unknown closed-loop system dynamics is realized.Using the learned knowledge to perform the same control task can avoid the retraining of the neural network.(4)Based on the trajectory tracking control for the single state-constrained wheeled mobile robot,the cooperative control of multi-mobile robots is further investigated.An adaptive neural network controller is designed using cooperative deterministic learning theory to guarantee each mobile robot individual tracking its own desired trajectory.Simultaneously,mobile robots exchange their weight information online via network communication,so as to achieve locallyaccurate identification of uncertain nonlinear dynamics on the basis of learning knowledge.The knowledge of cooperative learning can be stored and reused for robots to perform the same collaborative control tasks and the trained neural network also has good generalization capability when performing the control task over a domain consisting of the union of tracking orbits.The controllers designed above are all simulated and verified with the tracking control of wheeled mobile robots.Simulation results show that the control scheme proposed in this paper is effective and feasible.