Visual Servo Control of Robot Arm Based on Image Features
In recent years, with the increasing amount of research being done in the machine vision field and other related areas, the visual servo control of the robotic arm based on image information not only strengthens the diversity of information obtained by this device, but also expands its space, cognition and adaptability. It also improves the robotic arm’s ability to make precise identifications and execute fine operations. Processing image information and making good motion decisions are two of the skills that robotic arms need to master, as well as robotic arm vision. The image-based visual servo system does not require camera calibration, and can complete the motion control of the robotic arm through the image information in the camera plane. The control structure is relatively simple, so the current visual servo system has become a research focus. The online identification of
the image Jacobian matrix is also studied in order to obtain more accurate Jacobian matrix values in each iteration process of the image-based visual servo control system. In this paper, two different control algorithms are proposed for visual servoing. These two algorithms are based on the Kalman filter and neural network respectively, which can omit the image depth information required in the calculation of a visual servo system and reduce the extra overhead. In response to this, this paper proposes a hybrid kernel online sequence extreme learning machine (MIXEDKOSELM) based on hybrid kernel and online sequence learning. The correction of the error in the Kalman filter algorithm greatly improves the performance of the image-based
visual servo (IBVS) control system. This KF-MIXEDKOSELM-IBVS-based visual servo control method does not need camera parameters in the servo process, and is more robust against disturbance errors and noise statistical errors. The training results and test results of the MIXEDKOSELM algorithm on the abalone dataset were analyzed. It was concluded that the training results of two rbfs had MAE of 0.600 and TIMR of 11.12, and the test results had MAE of 0.579 and TIMR of 0.467. By comparing with other algorithms, it can be shown that the two rbf structures of the MIXEDKOSELM algorithm had obvious advantages in the testing process compared with other algorithms.
Keywords: vision servo; robotic arm; servo system; Kalman filter
Z. Ma, "Visual Servo Control of Robot Arm Based on Image Features",
Engineering Intelligent Systems, vol. 31 no. 6, pp. 501-511, 2023.