An Improved Model-Free Adaptive Control And Application In T Trajectory Tracking Of NAO Robot

Robot,as a typical representative containing various technologies,has high technology content and difficulty in research,which reflects the national technical strength and development level.Due to the high similarity with human beings,they can replace people to accomplish many difficult tasks,which makes a passion for many researchers.And NAO biped humanoid robot is widely welcomed because of its multi-degree of freedom,flexible action and high intelligence.However,the characteristics of its multi-degree of freedom and complex structure increase the difficulty in control,making precise trajectory tracking during walking become an important and difficult research.The complex trajectory tracking control of NAO robot is to reduce the trajectory tracking error and complete the trajectory tracking control task by designing a reasonable controller.However,the high complexity of the robot structure makes it difficult to obtain its accurate mathematical model,which brings many essential difficulties to the application of model-based control theory and method.Model-Free Adaptive Control(MFAC),as a typical representative of data-driven control method,has been successfully applied in many practical systems because of its strong adaptability,easy engineering implementation and no need for mechanism model during designing controller.Therefore,it is of great significance to apply MFAC method to the trajectory tracking problem of NAO robots.The main contributions of this thesis are summarized as follows:(1)The tracking control problem of NAO robot is described mathematically,and the desired trajectory in the trajectory tracking control of NAO robot is obtained by using A*algorithm,when the environment map and the starting and ending points are known.(2)For the disadvantage that the MFAC method only utilizes the online data in the control process,which makes the historical data be discarded,based on the idea of combining the controller dynamic linearization based MFAC method with the lazy learning method,two improved MFAC algorithms are proposed,i.e.two equivalent dynamic linearization of MFAC methods based on the controller compact format dynamic linearization and the controller partial format dynamic linearization,in which lazy learning method realizes the updating of controller parameters by synthetically utilizing on-line and off-line data of the plant.(3)First,the numerical simulation,conducted by the improved MFAC method,shows the advantages of the proposed method in tracking control of nonlinear systems by analysis from different aspects.Then the effectiveness of the improved MFAC method for trajectory tracking control of NAO robot is verified by comparison in simulation results of four control methods,in which the input and output data are provided by the dynamic model of NAO robot.Finally,Python programs are compiled to complete NAO robot's trajectory tracking control experiment,and the proposed method is compared with other data-driven control methods in three experimental scenarios,which proves the effectiveness and superiority of the proposed method.