Dead reckoning error compensation for the Xavier robot -------------------------------------- Henry Schneiderman Weh7220 25thJan95 A method has been developed and implemented for correcting the dead reckoning error in orientation on the Xavier robot. A linear model is hypothesized expressing the dead reckoning orientation error as a linear function of a set of state variables of the robot. The parameters of this model are estimated empirically by computing the least squares fit of this model to the dead reckoning error. The dead reckoning error is measured as the difference between the dead reckoning measurements of orientation and the sonar measurements of orientation. The estimated model has then been used to compensate for dead reckoning error in three experiments. In two of these experiments, compensation was nearly perfect. In the other experiment compensation achieved improvement, but not as accurately as in the other two experiments. The results also indicated that the dead reckoning error in orientation occurs primarily during translation of the robot.