Generalized architecture for simultaneous localization, auto-calibration, and map-building

This paper discusses the design of a very general architectural framework for navigation and tracking systems that fuse dead-reckoning sensors (e.g. inertial or encoders) with environment-referenced sensors, such as ultrasonic, optical, magnetic, or RF sensors. The framework enables systems that simultaneously track themselves, construct a map of landmarks in the environment, and calibrate sensor intrinsic and extrinsic parameters. The goals of the architecture are to permit easy configuration of numerous sensor combinations including IMUs, GPS, range sensors, inside-out bearing sensors, outside-in bearing sensors, etc., and to provide compatibility with multiple sensor networking standards, distributed sensor fusion algorithms, and implementation strategies. A decentralized Kalman filter based on Carlson's federated filter algorithm (1990) is used to decouple the auto-mapping, auto-calibration and navigation filters to produce a more flexible and modular architecture.