engleski

Simultaneous localization and mapping algorithm based on exactly sparse delayed state filter implemented using Lie groups

A prominent example of an application where the need for the computational benefits of the IF and the geometric accuracy of Lie groups arises is the simultaneous localization and mapping (SLAM). SLAM is of great practical importance in many robotic and autonomous system applications. Therefore, the extended information filter (EIF) is often employed and widely accepted for SLAM and has reached its zenith with sparsification approaches resulting with sparse EIF (SEIF) and exactly sparse delayed-state filter (ESDF). Here we propose a new state estimation algorithm called the exactly sparse delayed state filter on Lie groups (LG-ESDSF). The proposed filter is inspired by the filtering approaches on Lie groups and exhibits the advantages of the exactly sparse delayed state filter approach with regard to multi-sensor/multi-feature update and decentralization, while keeping the accuracy of stochastic inference on Lie groups. We present the theoretical development and demonstrate its performance on several examples. The results show that the filter achieves higher performance consistency and smaller error by executing SLAM directly on the Lie group.

SLAM, ESDSF, Lie groups

nije evidentirano