Modeling And Control Of Three-link Gymnastic Robot On A Horizontal Bar

Gymnastics robot is a nonlinear, non-integrity, strong coupling, mufti-state, underactuated system, and placed in the stable region is a naturally unstable system. Control of such systems in many areas that can reflect the key issues. Large numbers of domestic and foreign scholars engaged in gymnastics robot control.Although the simulation platform has gotten a good result, and there are many teams that have successfully controlled,in fact, the physical platform of success is so few. The key problem is that a large number of research teams are in the ideal simulation platform of gymnastics robot control, so they make no real sense of achievement and can not be used with the physical platform.This article discusses the "Quasi-equivalent Modeling" modeling approach, combined with the Lagrangian modeling method to develop an accurate model system for the control of late and lay a good foundation."Quasi-equivalent Modeling" is a method of estimating the system structure and the scope of the model parameters according to the system input and output response in the main characteristics (such as: delay, rise time, etc.), and then using search algorithm to adjust model parameters and ensure the deviation between the model and actual be in the allowed Range. This article firstly established a system of moment equations for the input of therigid body dynamics by using Lagrangian modeling method . Then analysed the three joint robot model gymnastics through Mechanism and Data. Got the robot drive system block diagram, sorting and simplified by the motor mechanism model,according to the principle of operation of the robot and control structure. However, the model structure is complex, not easy to control analysis.Further, the system has been reduced to a simple approximate model instead of mechanism model by employing the method of"quasi-equivalent"modeling. The pure delay link is determined by the step response curve, and the model order is determined by bode diagram of acceleration driving model which is got from frequency response method. Finally, the dynamic model of rigid body is simplified, and embeds the equivalent model of motor. The dynamic model which takes acceleration as input is established. The improved genetic algorithm has been indicated to identify model parameters. The results of experiments proved that not only the model precision is high similar in pre and post simplification, but also the deviation is low between simplified model and object platform. The simplified model is effective.Using multi-mode control idea of human-simulated intelligent control, the Swing-Up and inverted process of three-link acrobat is divided into multiple modes according to timing sequence, and every mode adopt for different controller. So, complicated control process can convert to some simple controls. Balance control of acrobat is the most critical and hard control stage. This paper use LQR controller, and use the improved genetic algorithm for searching optimal parameters. But the results of experiments show that the effects of using LQR controller on simulate platform build by accurate model individually are not well since stability region is too narrow.