UKF-Based Sensor Fusion Method for Position Estimation of a 2-DOF Rope Driven Robot

Abstract

In this study, the unscented Kalman filter-based method was introduced as a new technique for position estimation of the two-degree-of-freedom façade cleaning robot known as the Dual Ascender Robot (DAR). While other façade cleaning robots use a winch, the DAR uses an ascender, resulting in rope slip inside the ascender. Rope slip does easily cause errors in length data, so DARs cannot be easily controlled based on length data as in the case of most façade cleaning robots. Therefore, the DARs estimate the length data and use it through position estimation to overcome the rope slip for control. DARs use a rope length-based sensor fusion method for position estimation. This method employs position data based on both length data and angle data to estimate the position; however, it is difficult to use for long periods of time owing to the increased error that accumulates with time. Therefore, the use of position data based on angle data is proposed herein via application of the unscented Kalman filter. This unscented Kalman filter-based method is tested to confirm that the positional estimation performance is improved relative to that achieved via the previously used method. The performance improvements are compared in terms of accuracy and repeatability using the double ball bar method, and the errors in accuracy and repeatability are found to be reduced by approximately 2-3 times. INDEX TERMS Dual ascender robot, façade cleaning robot, sensor fusion, unscented Kalman filter, position estimation, IMU sensor.

Data Sensor Fusion for Surveillance Applications: Evaluation of Extended Kalman Filter vs. Unscented Kalman Filter

“…The UKF filter uses the UT (Unscented transformation) to improve the estimates of the first moments of a first random variable [9][10][11] by propagating a second Gaussian random variable through a nonlinear transformation [12]. This allows the UKF not only to have a higher convergence speed of the filter but also to always guarantee this convergence.…”

“…This allows the UKF not only to have a higher convergence speed of the filter but also to always guarantee this convergence. In order to take the CTRV model to three dimensions, the following assumptions were taken into account: Having x k , y k state vectors, u k work variables, v k noise associated with the process, and n k noise associated with the measurements for an instant in time k [11,13], the UKF filter starts from the following model:…”

A Survey on Techniques and Applications of Window-Cleaning Robots

“…To reduce this error, a sensor fusion method combining the rope length and angle sensing data with the weight factors was proposed. Through simulation and experimental study, it is verified that the sensor fusion method offered good anti-rope slip performance, and improved the stability (with about 4 mm error) under the sliding condition of 20 m × 20 m. To further improve the position estimation of the dual-ascender robot, Choi et al proposed a position estimation method based on the Kalman filter method for the rope angle data [93]. Experimental results of the accuracy and repeatability tests showed that the error of the sensor fusion method can be reduced by 2-3 times.…”

Ultra-wideband indoor localization method based on Kalman filtering and Taylor algorithm

“…In the actual working environment, the localization accuracy is affected by the NLOS error [9] . The Kalman filter-based data fusion algorithm is widely used [10] [11] , which can effectively suppress the NLOS error and achieve high localization accuracy in the complex environment. By analyzing the NLOS error and systematic error, this paper proposes a method to reduce the non-line-of-sight error and systematic error, using the coordinates of the solution after Kalman filtering as the initial point of the Taylor series expansion, and calculating the optimal value with the weighted least squares method to achieve the precise positioning of the tag.…”