Search

This paper proposes a robust image stabilization system for a mobile robot using Extended Kalman Filter (EKF). Though image information is one of the most efficient data for robot navigation, it is subject to noise which results from internal vibration as well as external factors such as uneven terrain, stairs, or marshy surface. The vibration of camera deteriorates the definition of image by destroying image sharpness, which seriously prevents mobile robots from recognizing their environment for navigation. In this paper, inclinometer was used to measure the vibration angle of the camera system mounted on the robot to obtain a reliable image by compensating for the angle of the camera shake caused by vibration. In addition angle prediction by using the EKF enhances responsibility of image analysis for real time performance. The Experimental results show effectiveness of the proposed system to compensate for the blurring of the images. [ABSTRACT FROM AUTHOR]

For pose estimation of the mobile robots repairing the hydraulic turbine blades, due to almost full pose range, there is the problem of degenerated pose matrix and singular solution of pose angles. A better approach to measure the pose errors of mobile robots using quaternion is suggested, called One-Rotational Pose Error (ORPE). Based on ORPE, the accumulated errors resulted from long-term use of the low-precision inertial system are studied, and the inclinometers are used to adjust the pose estimation. Theoretically, only inclinometers cannot adjust all the pose angles or the pose matrix. However, the Frobenius norm of the matrix or ORPE can be used to give the best adjustment. This paper deduces an interesting result, and simulation results not only testify the validity of the approaches mentioned above, but also show that the combination of low-precision inertial system and low-precision inclinometers can give rise to relatively high precision pose estimation. [ABSTRACT FROM AUTHOR]

In the shield tunneling construction, the inclinometer is usually utilized to measure the pitching angle and rolling angle of the shield machine in real-time. However, the nonlinearity and the temperature characteristic of the inclinometer always result in large measurement error. In order to improve its measurement accuracy, the research on compensating the nonlinear error and the temperature drift error based on BP neural network is presented in this paper. The characteristic of the inclinometer is studied by experiment at first and then its inverse model is built using BP neural network and trained with an amount of experimental data; finally the trained model is used to compensate the measurement error. The experimental results verify that the proposed compensation method can improve the measurement accuracy of the inclinometer greatly by correcting the nonlinearity and eliminating the influence of temperature. [ABSTRACT FROM AUTHOR]