Mechanical arm impedance learning control method based on width radial basis function neural network

Abstract

The invention discloses a mechanical arm impedance learning control method based on a width radial basis function neural network. The method comprises the following steps that a kinematic model of a mechanical arm is established according to the structure of the mechanical arm; based on a Lagrange equation and a kinematic model, a kinetic model of the mechanical arm is established in a task space; an expected task space regression trajectory model and a second-order impedance model are established; a width radial basis function neural network is constructed to realize dynamic adjustment of neural network nodes; a width radial basis function neural network is combined with a second-order impedance model to construct an adaptive neural network impedance controller; empirical knowledge is obtained based on the determined learning theory, and a constant neural network impedance controller is constructed. According to the method, accurate impedance control of interaction between the mechanical arm and the environment under the condition of unknown dynamic information is effectively achieved, the real-time performance of a control system is improved, and a novel safe and reliable method is provided for the situation that the mechanical arm repeatedly interacts with the environment.

Description

technical field [0001] The invention relates to the technical field of safe and compliant control of a manipulator, in particular to a method for learning and controlling the impedance of a manipulator based on a width radial basis neural network. Background technique [0002] With the rapid development of science and technology, robotic arms have been widely used in industry and service industries, and the control tasks they face are becoming increasingly complex. In the face of many work situations, such as manipulator grinding, assembly work, rehabilitation medical work, human-machine collaborative work, etc., the traditional position control of the manipulator can no longer meet the control requirements in this area, and the compliant control of its force often needs to be considered . In the compliant control algorithm, impedance control has the characteristics of strong anti-disturbance ability and easy force control of the mechanical arm by incorporating force and po...

AI Technical Summary

Problems solved by technology

In the research and application of the impedance control of the manipulator, the calculation of the control torque of the manipulator often requires an accurate dynamic model to achieve control accuracy, and the manipulator system has modeling uncertainties due to factors such as friction and damping. Affect the compliance performance of the robotic arm

In order to overcome the problem of control performance degradation caused by modeling uncertainty, the neural network is usually used to identify the unknown dynamics of the manipulator, but the traditional adaptive neural network controller needs to be adjusted online when performing the same or similar control tasks for the manipulator. Weights are used to re-identify unknown dynamics, which makes the control scheme occupy a large amount of computing resources, take a long time, and is difficult to implement

In addition, the structural parameters of the traditional radial basis neural network usually rely on the designer's experience and trial and error to select, which is subjectively biased and inefficient. When a higher neural network approximation accuracy is required, it is often necessary to select too many parameters. central point, which further increases the computational load of the neural network and affects the real-time performance of the control system

Images