Study On Dynamic Characteristics And Related Technology Based On Bionic Algorithm For Robot Wrist Force Sensor

A new kind of robot multi-dimensional wrist force sensor is chosen as studying subject. Some issues on multi-dimensional wrist force sensor, including output force signal de-noising methods, calibration methods, dynamic modeling methods, dynamic compensation methods, dynamic characteristic interacting with environment and dynamic modeling compensation device design etc., are studied in details. Bionic algorithm (BA) is applied to the proposed research. The main work is organized as follows:1. Some progress about robot multi-dimensional wrist force sensor is reviewed and some problems existed in multi-dimensional wrist force sensor research are presented. The foundational principle of BA is introduced. It is pointed out that it is key problem that enhances the dynamic modeling accuracy by using novel and effective algorithms and improves the real time and practicability of dynamic compensation device by using new principles, new methods and new devices for robot multi-dimensional wrist force sensor.2. A new kind of robot multi-dimensional wrist force sensor is introduced and its structure, working principle and signal acquiring method etc. are discussed in details. A finite element analysis for the elastic body of the sensor is made by means of ANSYS. The result validates the rationality of the design.3. The couple of multi-dimensional wrist force sensor is one major factor of limiting measurement precision. A new method to multi-dimensional wrist force sensor calibration based on RBF neural network is described for the shortcoming of conventional method. Apart from the RBF neural network based a new multi-dimensional wrist force sensor calibration methodology. The comparison of the calibration results is also presented. These results show that the proposed RBF neural network method has high precision compared with least square method and has fast training speed compared with BP neural network. The new method is valuable for practical applications.4. During robot wrist force sensor works, the signals are inevitably influenced by stochastic noises, which have poor effects on its measure andcontrol precision. In order to overcoming the shortcoming of low pass filter and FFT/IFFT, the theory of multi-wavelet is analyzed and applied to process the output force signal for robot wrist force sensor, and the method of soft threshold is adopted during the data process. The comparison of the de-noising results is also presented. The experimental results show that the proposed method is effective for eliminating the effect of stochastic noises, but the performance of the new method excels that of the conventional method.5. A new dynamic modeling and compensation method is presented based on improved genetic algorithm (IGA) where the operator of crossover and mutation is improved and function link artificial neural networks (FLANN) and the dynamic modeling and compensation principle and algorithms are introduced for a new kind of robot multi-dimensional wrist force sensor. In this method, the dynamic model and compensation model of wrist force sensor can be set up according to measurement data of the dynamic calibration, where the dynamic model and compensation model parameters are trained by improved genetic neural network. So the method remains the global searching ability of GA and the simple structure and good robustness and self-learning ability of FLANN. The results show that the proposed new dynamic modeling and compensation method can overcome the shortcomings of FLANN, such as easy convergence to the local minimum points and slow speed in training, and has the advantages of fast training process, good real time, good global searching ability, high precision, and easy realization of dynamic compensation device.6. The dynamic characteristic of multi-dimensional wrist force sensor interacting with environment is investigated. The relationship between natural frequency and environment parameters is studied. The investigation is of important value for realization of dynamic compensation. In order to sense the environment, it is necessary to building model. For this reason, the dynamic model of operating environment is further researched. Firstly, the building method of dynamic model of operating environment in robot system based on RBF neural network is presented, where the mechanism and the algorithm of building model are elaborated. Secondly, the building method of dynamic model of operatingenvironment in robot system based on wavelet neural network (WNN) is presented, where geometrical structure of the network is analyzed and the method of network parameters training and initialization is given. The weights of network, scale factor and displacement factor are studied by the steepest descent method, and network parameters initialization integrates with the wavelet type, time-frequency parameters of wavelet and the training samples. The comparison of the modeling methods is also presented. The results show that the proposed methods provide better approximation ability and higher precision and faster training speed than the BP neural network when used in approaching nonlinear systems and provide the powerful tools for the dynamic modeling of operating environment in robot system. The research is of important significance for design of control architecture and control algorithm of robot system.7. A new adaptive dynamic compensating devices are designed based on charge transfer devices (CTD) for robot multi-dimensional wrist force sensor, where the frequency characteristic is adjustable. The compensating principle and circuit design are described. The theoretical analysis and experiment result show that the generality and real time of compensating devices are improved. The design is of important value for realization of the dynamic compensation device of robot multi-dimensional wrist force sensor.