Research On Multi-objective Traijectory Optimization And Motion Control For Serial Robot Manipulators

With the remarkable improvement of domestic industrial automation, robot manipulator installations are increasing greatly in our country, the robot manipulators have become standard equipments and have been widely used in the industrial fields. Experts predict that the amount of domestic robot manipulators will be increased by 25% every year. However, some foreign companies, such as ABB, FANUC, YASKAWA, MOTOMAN, KUKA, AdeptTechnology, COMAU, etc. have an overwhelmingly large share of our market while domestic companies have less than 1%. Moreover, there is a big gap of these likewise products between China and other countries on the performance such as speed, accuracy, reliability etc. The technology of robot manipulator is a cross-disciplinary integration, including mechanical, electronic, computer, control, information, artificial intelligence, bionics and other of high-tech field. The theoretical research of robot manipulators is very active. Hence, based on the domestic technological foundation, deep research on the general technology and the key technology of robot manipulators especially the research on the high-speed and high-precision technology can not only improve the theory research level of robot, but also will promote the overall technology of robot manipulators in China.This dissertation addresses the study of our self-made SCARA (Selective Compliance Assembly Robot Arm) robot. Based on the knowledge of many disciplinary, such as signal analysis, modern control, system identification, nonlinear system and so on, with the help of combination of theoretical analysis, numerical simulation and experimental validation, high-speed and high-precision motion of robot manipulators and its related area are studied. For the system realization, based on the domestic existing technology, the robot manipulator design use all the domestic devices, and in order to guarantee the overall performance of the robot manipulator, The system design is optimized and advanced control algorithms are adopted to compensate for the lower performance of the devices themselves. In the electric system, ARM9 platform with Linux embedded are adopted in the design of main controller to realize human-computer interaction and motion planning. DSP+CPLD platform is used for the motion controller. These will help to reduce the cost and improve the stability of the system. In the design of mechanical structure, domestic harmonic reducer has been chosen, and corresponding transmission, bending and centering device are designed. With the help of advanced motion control algorithms, the flexibility and nonlinear friction factors of harmonic reducer are compensated. To obtain high-speed and high-precision motion performance of robot manipulator, some theory such as motion planning, motion control and flexible joints control, etc. are discussed. The maximum speed of each joint has reached more than 90% of the rated speed of robot manipulator, the tracking accuracy still reach less than 10 times of the sensor precision. Eventually the tracking error on each joint is less than 0.005°for multi-axis motion control and even less than 0.003°for uniaxial motion control. Such work is a exploration in realizing high-performance localized robot manipulator on the key technologies and the practice of overall unit.This dissertation is divided into seven chapters, the main contents are as follows:The first chapter introduces the research background, research topics related to the foreign advanced robot products and industrial robot technology, analyzes the latest progress of domestic robot manipulators and the present situation of the technology and industry. Literatures of the robot theories, including structural optimization, control system, trajectory planning, motion control and flexible joints, are reviewed. At last the support and the background of this research are clarified.In chapterⅡ, the system realization and key technologies of the self-made SCARA robot are introduced, including the structure design, control system and drive system, etc. In the structure design, harmonic reducer components are adopted, corresponding transmission, bending and centering device are designed. In the control system, Teaching and planning system based on ARM9 platform is designed, and corresponding human-machine interface and motion planning algorithm are developed under Linux system, the function of robot demonstration and teaching planning are completed. Motion interpolation, high-speed and high-accuracy motion control algorithm are implemented on motion controller based on DSP. At last low cost prototype design are completed.In chapterⅢ, the multi-objective trajectory optimization for manipulators is studied. The high degree B-splines are adopted to construct a continuous trajectory in joint space which guarantees the continuity of motion, as well as free configuration of motion parameters of starts and stops. According to the convex-hull property, kinematic constraints are transformed to the constraints on control points of B-splines. With minimum time, smooth trajectory, lowest energy consumption and the least torque variation as our optimizing objectives, firstly, the improved non-dominated sorting genetic algorithm-Ⅱ(NSGA-Ⅱ)is applied, then the Pareto optimization solution set is obtained, which provided many choices for users. The results on the general six degrees of freedom manipulator show that compared with the single-objective optimizations, the multi-objective optimization algorithm is better.In chapter IV, some model free controllers are studied for trajectory tracking of robot manipulators. Firstly, a supervisory adaptive PID controller is proposed based on the traditional PID controller. Then the adaptive PID algorithm is designed to minimize the accessibility conditions to precipitate the sliding mode occur. At the same time, a supervisory adaptive RBF neural network controller is presented, the RBF neural network approximates the ideal controller and the supervisory controller guarantees the system stability. Then a robust nonlinear PID output feedback controller is designed. The PD part stabilizes the system, the nonlinear integration part is designed to improve the dynamic response and the robustness. A linear observer is proposed to estimate the actual joint velocity to avoid the high frequency disturbance generated by differential computing the position value. The experimental and simulation results show that the controller can achieve the favorable tracking performance without the model parameter information.In chapter V, the research is focused on the parameter identification and adaptive robust control based on the model information. Firstly, the dynamic model of the robot manipulator is constructed and linearized. Then in consideration of the dynamic constraints, the optimized exciting trajectory is designed to excite the dynamic characteristics of the manipulator, so as to reduce the identification error, and the maximum likelihood method is used to get the precise parameters. It's shown that the proposed technique is effective according to the simulation and experimental results. A controller based on the model is proposed, position-dependent friction is expressed as the combining sine and cosine function with unknown coefficient, based on the desired compensation adaptive robust control, an improved adaptive robust control (IARC) is proposed. The controller replaces the integral parametric estimations law with the proportional-integral parametric estimations law considering the "slow" learning rate to suppress the high-frequency noise. The experimental results on the single-joint and multi-joint manipulator show that the controller can achieve good tracking performance.In Chapter VI, dynamic surface backstepping control and saturated fuzzy dynamic surface backstepping control are proposed to deal with the flexible joint robot derived by harmonic gear reducer. Firstly, the traditional backstepping controller is designed for the flexible joint, and then the dynamic surface technique is introduced to design the dynamic surface backstepping controller. A first-order low-pass filter is proposed to approximate the virtual controller generated by backstepping so as to avoid the derivation of the controller. Thus the problem of 'explosion of complexity'in backstepping design procedure is solved. Based on the controller a saturated fuzzy dynamic surface backstepping controller is proposed to deal with the problem of limited torque, fuzzy system is used to approach the saturation nonlinearity to improve tracking precision under the limited torque.In chapter VII, the main work, results and innovations of the thesis are summarized. And the future research is also prospected.