| Hydraulic manipulator has the characteristics of strong load capacity,compact structure and rapid response,and is widely used in various fields of industry.Taking the multi-degree-of-freedom manipulator based on the hydraulic swing cylinder joint as the research object,the control system were studied in detail,and a hydraulic manipulator control system with low cost and non-servo solenoid directional valve as the control element was designed.Aiming at the undesirable control characteristics of the solenoid directional valve,the positioning control of the hydraulic manipulator is studied,and a more complete control theory and method are put forward.The specific research content is divided into the following three aspects:(1)Taking the designed multi-degree-of-freedom hydraulic manipulator as the research object,the kinematic analysis of its overall structure was carried out with the improved D-H method,and the forward and inverse kinematic solutions were solved in detail.Aiming at the hydraulic control system with solenoid directional valve as the core,the motion mode of positioning control is selected and the corresponding motion learning strategy is developed.According to the motion control requirements of the manipulator,the characteristics of the whole control system are extracted and analyzed,and finally the mathematical model describing the input and output characteristics of the whole control system is derived.(2)According to the developed learning strategy,the corresponding control system is designed.Starting from the hardware and software aspects,it has the functions of man-machine interaction,motion control and sensing detection,etc.The main objective is to achieve the artificial teaching of the manipulator and the collection of teaching information.The operator can drive each joint of the manipulator arm to complete various teaching movements through the upper computer or the main controller,and the related setting parameters and the motion information of the sensor are saved as teaching information to facilitate future teaching reproduction and provide training samples for the regression model.(3)The machine learning algorithm is used to carry out regression analysis on the collected training samples to establish the prediction model of manipulator control.The control system can output the control command corresponding to the action according to the established model,and the final execution result will be feedback by the sensor.The cases with substandard precision will be saved as new training samples,so as to continuously improve the prediction model and improve the accuracy of positioning control of the manipulator. |
PDF Full Text Request