State estimation and parameter identification of flexible manipulators based on visual sensor and virtual joint model

Abstract

When we control a flexible manipulator based on a dynamic model, we need to estimate the state variables with accuracy. In the case of rigid manipulators, the state variables consist of joint angles and their velocities which are measured easily by encoders, potentiometers, tachometers, and so on. In the case of flexible manipulators, however, the state variables also include elastic deformations and their velocities due to flexibility. Here we focus on visual sensor to measure distributed state variables of flexible manipulators. First, we discuss a distributed state estimation method which uses discrete position information of markers attached on flexible links. Second, when we apply the virtual joint model to flexible manipulators, elastic deformations are expressed by virtual passive joint angles which are lumped. So we have to correspond the real elastic deformations to the virtual joint angles. We also discuss this transformation. Third, when we estimate the lumped state variables by the proposed method, it is possible to identify physical parameters of the dynamic model. Finally, some experiments were carried out to verify the effectiveness of the proposed method.