A method to convert floating to fixed-point EKF-SLAM for embedded robotics

The Extended Kalman Filter (EKF) is one of the most efficient algorithms to address the problem of Simultaneous Localization And Mapping (SLAM) in the area of autonomous mobile robots. The EKF simultaneously estimates a model of the environment (map) and the position of a robot based on sensor information. The EKF for SLAM is usually implemented using floating-point data representation demanding high computational processing power, mainly when the processing is performed online during the environment exploration. In this paper, we propose a method to automatically estimate the bit-range of the EKF variables to mitigate its implementation using only fixed-point representation. In this method is presented a model to monitor the algorithm stability, a procedure to compute the bit range of each variable and a first effort to analyze the maximum acceptable system error. The proposed system can be applied to reduce the overall system cost and power consumption, specially in SLAM applications for embedded mobile robots.