Relative observation for multi-robot collaborative localisation based on multi-source signals

This paper describes a synthesising method for multi-robot collaborative localisation. A distributed extended Kalman filter (EKF) based on robot odometry and external North Star signals for data fusion is first designed for the localisation of individuals in the robot group. Relying on relative observation by infrared sensors and gyroscopes mounted on robots, and the 'uncertainty volume' strategy, the positions estimated by EKFs are further corrected for precising the localisation process. The localisation accuracy based on different sensing regimes is tested. Sensor correlations and uncertainties are analysed for predicting error propagation and to accommodate sensing deviations. The multi-source signals are then synthesised for the collaborative localisation for a multi-robot system without introducing excessive computation. Finally, this work is verified by both simulation and experiments with real robots, i.e. the Festo Robotinos under different scenarios.