КОМБІНОВАНИЙ МЕТОД ВИЗНАЧЕННЯ ПОТОЧНОГО ЗНАЧЕННЯ КООРДИНАТ НАВІГАЦІЙНОЮ СИСТЕМОЮ З ФІЛЬТРОМ ЧАСТИН РАО-БЛЕКВЕЛЛА

Abstract

Modern trends in the development of information technology and robotics, sets before us the task of adapting obsolete navigation systems of military equipment to the requirements of our time. Existing navigation systems are widely used in both military and civilian spheres The analysis of existing navigation systems showed a number of negative problems, which became the reason for work to create a navigation system that could meet the following requirements: low financial costs for production, mobility, small overall dimensions, accuracy in determining coordinates and the most important - autonomy. The most effective methods for positioning autonomous moving objects is the method for calculating the coordinates of the EKF-SLAM with the Rao-Blackwell parts filter. This article presents a combined method for determining the current coordinate value by a navigation system with a Rao-Blackwell parts filter. The Rao-Blackwell filter for mapping progressively processes observation and odometer evidence as they become available. This process is accomplished by resuming a set of samples that represent the posterior on the map and the vehicle’s path. The proposed way of modifying the well-known mathematical relations of Kalman filters from the point of view of their adaptation to the peculiarities of algorithmic and software implementation in onboard computers provides savings in onboard computer memory and reduces the required computing resource. It is noticed that the algorithms for the implementation of SLAM navigation are changed in the proposed way using a smaller number of particles than methods based only on the frequency filter. The error in the initial calculation of the coordinates of landmarks is minimized and does not accumulate over time in a mathematical sense. Keywords: Autonomous mobile object, Simultaneous Localization And Mapping, EKF-SLAM, Gmapping, Rao-Blackwell particle filter, covariance, parameter prediction, positioning, orientation algorithm, Kalman filter.