Joint Friction Analysis And Low-speed High-precision Motion Control Of Multi-DOF Serial Robots

Multi-DOF serial robots are widely used in fields such as industrial automation, medical, aerospace, marine. In spite of the seeming mature of industrial robot technology after decades of development, high precision motion control in extreme conditions is still a problem, e.g., moving at extreme low or high speed. So it is difficult for robots to meet the requirements of applications which need low-speed high-precision or high-speed high-precision movement. This dissertation is dedicated to study on the low-speed high-precision motion control of multi-DOF serial robots. After the robot structure determined, we try to improve the low-speed performance mainly through following ways:model and compensate the joint friction, estimate the joint velocity accurately, improve the control algorithm design.Firstly, we introduced the background of the research and conducted a comprehensive analysis of factors affecting the low-speed motion performance, including structure design, sensor resolution, nonlinear phenomenon, controller design, etc. Structural design and optimization, joint velocity estimation, joint friction modeling and compensation, as well as motion control strategy are currently the major research directions.Designing model-based control algorithm is an effective way to improve low-speed control precision. However, it needs to know the robot dynamic model and parameters. After the dynamic modeling, we studied on a systematic off-line dynamic parameter identification method conducted under closed-loop control in chapter Ⅱ. Since traditional Fourier series do not satisfy velocity and acceleration boundary conditions, we designed modified Fourier series as exciting trajectory. Then, to minimize the sensitivity to measurement noise, the coefficients of modified Fourier series were optimized according to the condition number criterion using genetic algorithm. In addition, considering the measurement noise effects, maximum likelihood estimation method was adopted to obtain accurate parameter estimates.Among the nonlinear phenomenon, joint friction is the main factor which causes control performance deterioration at low speed. Since RV reducer and harmonic reducer are widely used in robots, we studied on the friction of RV-drive-based joint and harmonic-drive-based joint in chapter Ⅲ. We mainly discussed the friction sources, measurement method, the friction characteristics, modeling and identification. For RV-drive-based joints with large inertia, we analyzed the friction parameter variation with different load torque. Then, based on the exponential model, we proposed an extended friction model which can describe the load torque effect. For harmonic-drive-based joints, the friction changes periodically with the joint angle. Based on the frequency domain analysis of the friction data using FFT, the main frequency components were determined. Then a friction model, which is the combination of the exponential model and sine/cosine functions, was involved to model the friction.In robot systems, high-quality joint velocity estimation is of great significance for friction compensation, full state feedback controller design, etc. Before controller design, it needs to design an efficient method for velocity estimation. At low speed, quantization error and measurement noise may cause large velocity estimation error. To deal with this problem, we discussed the use of nonlinear tracking differentiator to estimate joint velocity in chapter IV. The nonlinear tracking differentiator has advantages such as easy to use, simple parameter tuning procedure, independent of model, which make it well adapted to robot system.Based on the modeling and parameter identification in chapter Ⅱ and Ⅲ, we designed model-based controllers. For RV-drive-based joint, since the friction uncertainty, fixed compensation can not work well. To deal with this problem, we adopted adaptive control to learn the uncertainty. First, fuzzy logic system was involved to approximate the friction phenomenon, and linear model was derived for design of adaptive learning algorithm. Then, we proposed a fuzzy adaptive robust controller. For harmonic-drive-based joints, due to the cyclical changes of friction, it is difficult to obtain linear form, and adaptive control is not applicable. Since robust control can guarantee stable and performance with bounded uncertainty, we proposed a robust controller with friction compensation. For robots with multiple joints (>3), it is difficult to design model-based controller due to the model complexity, large calculation and difficulty of parameter identification. Controllers which have strong real-time, simple implementation and low dependence of the model are favored by researchers. We tried to design controllers with little dependence or independent of the model in chapter VI. Firstly, we designed a fuzzy PID controller with friction feedforward. In the controller, Mamdani fuzzy logic system was adopted to tune the controller parameters. Exponential model was used in the friction compensation. For applications with the friction model unknown, we proposed a sliding PID controller with tracking differentiator. As we known, the PID control has good steady-state accuracy. Sliding mode variable structure control has the features of fast response and strong robustness. The designed controller combines the advantages of both. Experimental and simulation studies were conducted. Two typical operating trajectories in robot laser cutting, i.e., small circle and waist-shaped hole, were tested.