Fuzzy Landmark Detection in Simultaneous Localisation and Mapping for Agricultural Robots

This paper deals with the problem of landmark detection for application of Simultaneous Localization and Mapping (SLAM) for agricultural robots. A new method of using a fuzzy rule base for fast evaluation of landmarks based on its measurement error, amount of dynamic nature, need for a landmark and evaluation time has been proposed. The output of the fuzzy system is a multiplier which represents the degree of uncertainty of the landmark. High values of uncertainty results in landmarks being discarded, average values results in revaluation, and smaller values result in acceptance with the error covariance being a product of the multiplier and the measured covariance. Selected landmarks with corresponding error covariance are used to update the Kalman Filter. Simulations show the effectiveness of the fuzzy system in improving the relative position uncertainty.