Towards a framework for airborne EO/IR surveillance. An introduction to simultaneous localisation and map building

The need for autonomous sensor data processing and sensor management in UAV surveillance and reconnaissance systems will increase. An important question is how to fuse sensor observations with prior information and navigation data, taking into account their associated uncertainties in an appropriate way. Another question is how to represent objects and landmarks in the environment in order to facilitate distributed data fusion and cooperative information gathering using multiple platforms. Simultaneous Localisation and Map Building (SLAM) is the process of concurrently building up a feature-based map of the environment and using this map to estimate the location of the platform. The platform starts in an uncertain or unknown location, with or without prior information about the environment. The position and orientation of the vehicle and the locations of landmarks (features) are estimated on-line with dead-reckoning and relative observations of the landmarks. A number of methods have been proposed for addressing the SLAM problem. In this report two methods are presented, EKF and SEIF. EKF SLAM is based on the well-known Extended Kalman Filter and SEIF (Sparse Extended Information Filter) SLAM is based on the information form of the EKF. In SLAM, the normalized information matrix is typically sparse and the SEIF method exploits this fact by approximating the information matrix. The resulting algorithm is much more computationally efficient than EKF SLAM. The- report ends with a discussion about future work on multi-vehicle SLAM with EO/IR-sensors.