Genetic Algorithmic Filter Approach to Mobile Robot Simultaneous Localization and Mapping

This paper presents a genetic algorithmic filter (GAF) approach to mobile robot simultaneous localization and mapping (SLAM). A Genetic algorithmic approach is used to solve the SLAM problem by concurrently optimizing appropriate cost functions defined over two sets of chromosome populations representing the robot pose and the environmental map. As such the methodology has the potential to produce globally consistent solutions to this highly non-linear, non-Gaussian SLAM problem. Similar to Monte-Carlo probabilistic localization, particles or chromosomes are used to represent the belief state of the robot and the environmental map. However, unlike in the case of a particle filter approach such as FastSLAM, the genetic algorithmic approach presented in the paper does not rely on environmental feature extraction and data association. Further, it is shown how the problems of sample impoverishment associated with re-sampling which is common in a particle filter approach can be systematically and effectively overcome through the application of a parallel variant of the genetic algorithm with associated operators. Simulation and experimental results are presented to demonstrate definite performance gains achievable through the use of GAF-SLAM and its potential to yield globally consistent SLAM results

I. Introduction

The Simultaneous Localization and Mapping (SLAM) problem, also known as Concurrent Mapping and Localization (CML) problem is often recognized in the robotics literature as one of the key challenges in building autonomous capabilities for mobile vehicles. The goal of an autonomous vehicle performing SLAM is to start from an unknown location in an unknown environment, and build a map of its environment incrementally by using the uncertain information extracted from its sensors, whilst simultaneously using that map to localize itself with respect to a reference coordinate frame and navigate in real-time. In SLAM the map is represented in one of several forms. A common approach is to use features extracted from exteroceptive sensors' data [1]. Another approach is to use a grid to represent the map [2].