Feedforward global/inertial sensor fusion algorithm for accurate global positioning of a mobile robot

This paper introduces a coordinate transform method for global/inertial sensor fusion minimizing modification of an existing control program of a mobile robot Most of GPS/INS sensor fusion algorithms use Kalman filters and modify the INS states by feedback loops. Because the structure of the proposed method has a feedforward filter, the proposed method has an advantage in case the user does not want to change an existing control program of mobile robot. The feedback type Kalman filter is designed so that the error between global position from GPS and odometry from INS converges to zero. Therefore the coordinate matching between the odometry and measured global position is not necessary in those approaches. However, in the feedforward structure described in this paper, the errors gradually increase over time. A coordinate transform method has been developed for dealing with the error. This method provides an easy way to make an add-on function without any changes in the existing functions of the control program of a mobile robot