| Modular robots are composed of modules that integrate power supply,computing system,sensors and actuators.These modules usually have standardized mechanical and electrical interfaces and can be assembled freely according to different task requirements.In addition,a control system is needed to ensure the stability and precise motion control of modular robots.This paper did the following research for modular robots:(1)Dynamic model of harmonic drive-based modular robots with joint torque sensingThis paper introduced the motion and torque transmission principle of the harmonic drive,which is an important component of the modular robot joint module,and established a mathematical model based on the flexibility of harmonic drive.The disturbance of torque transmission of harmonic drive and the measurement error of sensor are taken into account.Combined the idea of modularization,the dynamic model of modular robot system is established based on joint torque sensing technology.(2)Robot joint velocity estimation based on inverse hyperbolic sine tracing differential algorithmIn this paper,a novel method of joint velocity estimation for robots is studied based on local position information.Unlike the conventional methods that using the motor-side position measurements to obtain the joint velocity,this paper addresses the problem of joint velocity estimation that using the link-side absolute encoder data,so that the kinematic error caused by the mechanical transmission can be avoided effectively.Based on the inverse hyperbolic sine tracing differential algorithm,a velocity estimator is designed to obtain the high precision of the joint velocity and suppressing the measurement noise effectively.The stability of the proposed velocity estimator is proved by using the Lyapunov theory.Finally,experiments are performed to verify the effectiveness and advantage of the proposed velocity estimation method.(3)Robot joint torque estimation based on harmonic drive flexible model and double encoder measurementsIn order to achieve high control accuracy for modular robot systems,it is necessary to consider the compensation of the constraints and load force of the robot joints.Joint torque feedback technology is considered to be an effective way to solve this problem.In order to obtain joint torque without using torque sensor,a joint moment estimation method is introduced based on the improved nonlinear harmonic drive model.Considering the flexibility of wave generator and flexspline,the kinematics error of harmonic drive and the torque deviation,estimated the joint torque by only using the position measurement data acquired by the encoders at the motor side and the end of joint.Moreover,the applicability of this method is verified by experiments.(4)Robust neural network decentralized control method based on local information measurementDecentralized control method decomposes the complex dynamics model of robot into multiple coupled subsystems,and designed the local controller to compensate the uncertainties and coupling effects of the subsystem model simultaneously.The advantage of the method is reducing the complexity of the control system from the whole system level to the decentralized subsystem level,which greatly simplifies the complexity of the controller and effectively improves the computation speed,it also has strong fault-tolerant ability.In this paper,in the case of limited sensor measurements,using the estimated joint velocity and joint moment obtained from the previous research to realize decentralized control of modular robot only based on local joint information measurement.The controller also approximates and compensates the uncertainties of the robot model based on the neural network learning function,and achieves high-precision control effectiveness.Finally,the stability of the controller is proved by Lyapunov stability analysis method,and the effectiveness and advantages of the controller are verified by experiments. |
PDF Full Text Request