Robust localization algorithms for outdoor mobile robot

On the basis of Pioneer3-AT wheeled outdoor mobile robot, coder, gyro, compass and RTK-GPS were used for the robot localization. A new kind of federated Kalman filter was designed in which data from multiple sensors were fused stage by stage. Data of odometry and gyro were fused firstly. Then the fusion data were fused with data of compass. Data of RTK-GPS were fused finally. This filter could filter the fluctuation and compensate error accumulation of sensors and could achieve good precision of robot localization. In city environment, RTK-GPS would lose difference state frequently because of the shelter of buildings, so the precision of RTK-GPS was not steady. This filter could choose different measurement error covariance matrix by the precision of RTK-GPS, which made the filter adapt well to the change of RTK-GPS precision automatically, so the localization algorithm was robust. Experiment results show that robot can achieve the localization precision about 0.4 m steadily.