Ant system based state estimation approach to SLAM

This paper proposes an ant system based state estimation approach for simultaneous localization and mapping (SLAM) in the case of ambiguities both in the feature number and data correspondence. Inspired by the random finite sets (RFS) and its derivative, i.e., probability hypothesis density (PHD), an ant-PHD filtering is proposed to jointly estimate the locations and number of features, moreover, a fast moving ant estimator (F-MAE) is developed for estimating maneuvering vehicle trajectory. In contrast to the state-of-the-art approaches, our algorithm employs the artificial ants instead of simple particles to cluster around their favored regions through ants' positive feedback search mechanism, and also builds a seamless from the filter itself to implementation. Simulated results demonstrate the merits of the proposed approach, which outperforms both the Fast-SLAM and the PHD-SLAM by providing a more accurate map as well as an improved estimate accuracy of the vehicle's trajectory.

I. Introduction

Simultaneous localization and mapping (SLAM) in robotics is usually described as follows: a robot uses its self-contained sensors to measure the operating environment and incrementally build environment map, while concurrently determine its pose relative to the map [1]. The research of the SLAM technology has important theoretical significance and application value, which can provide a theoretical basis and technical support for the development of autonomous underwater robots, space exploration robots, unmanned aerial vehicles and other autonomous systems.