This thesis presents the progress of work toward creating a localization engine for a fleet of autonomous ground robots using ultra-wideband (UWB) beacons in concert with the robots’ on-board IMUs and encoders. The error observed from the uncorrected UWB ranges was quantified for line-of-sight conditions, and a calibration procedure was developed and applied for UWB range correction. Four localization approaches were implemented and tested on a ground robot in an outdoor environment. Dead reckoning was used as a localization baseline. An extended Kalman filter (EKF) using the manufacturer firmware’s calculated positions for the measurement input was implemented and resulted in a final pose error of 29 cm. A second EKF using calibrated ranges and least squares estimation (LSE) for state measurement was implemented and resulted in a final pose error of 26 cm. A third Kalman filter, using an experimental extension of a mesh relaxation technique for position estimate, was tested but found not to track the position of the robot well. Discussion of implementation experience and recommendations for future work are provided.
【 预 览 】
附件列表
Files
Size
Format
View
Localization for teams of autonomous ground vehicles