Analysis And Research On Blind Guiding Robot Localization And Orientation Based On Multi-Sensor Fusion

The multi-sensor fusion plays an important role in implementation of localization and orientation. It becomes a robotic research hotspot that this technology is applied to the blind guiding robot for low vision population. This paper focuses on the analysis and study of the localization and orientation based on multi-sensor fusion. And the perception system is hierarchically designed.The sensing strategies of ultrasonic sensors, electronic compass, odometer and RFID measurement is modeled on process layer. A double LANDMARC method for RFID ranging is proposed. Characteristics and error causes of sensor modules is analyzed. Error compensation methods are used to correct sensor noise data for more accurate integration resources.Data fusion layer methods are analyzed and studied in details. An unscented Kalman filter based on confidence matching is proposed. Firstly, input data confidence weight is given by confidence matching matrix. Then position and direction data of robot is unscented transformed to generate several Sigma sample points.After that, Sigma sample points and observation data is fused with the unscented Kalman filter for pose estimation to get more realistic information of the localization and orientation system. At last, three filters are compared by experiments on their practical application. The experiment result shows that the proposed unscented Kalman filter based on confidence matching completes tasks for the robot localization and orientation more accurately and efficiently.