Mutual Localization of Mobile Robotic Platforms Using Kalman Filtering

The ability of a mobile robotic platform to self-locate within its environment is needed to implement navigation functionalities. In situations where several platforms cooperate to jointly execute tasks, mutual platform localization is paramount to control the robot formation. This paper presents a real-time mutual localization system for multiple mobile platforms. The odometry system of each platform is used to calculate its movement. The updating of the mutual localization uses data sharing between platforms but does not require visual contact between them. An onboard omnidirectional vision system is used for landmark detection. While landmark locations are unknown, their positions within the field of view of the platforms are used for localization error correction, through variable-dimension extended Kalman filtering. The method is fast, robust, and provides accurate results with low-cost equipment