Sonar-Based Simultaneous Localization and Mapping Using a Neuro-Evolutionary Optimization

Abstract

This paper addresses a solution of simultaneous localization and mapping (SLAM) for sonar readings based on neuro-evolutionary optimization algorithm. In the past two decades, numerous studies have attempted to solve the SLAM problem using laser scanners and vision sensors. However, relatively little research has been carried out on a sonar-based SLAM algorithm, because the bearing accuracy and resolution of sonars are not enough to find consistent features for SLAM. The proposed algorithm in this paper solves the sonar-based SLAM as a global optimization problem using the cost function that represents the quality of a robot's trajectory in the world coordinate frame. In our algorithm, a neural network helps to estimate the robot's pose error accurately using sonar inputs at each position and the pose difference between two consecutive robot poses, and evolutionary programming is used to find the most suitable neural network. By way of learning and evolution, our algorithm does not need a prior assumption on the motion and sensor models, and therefore shows a robust performance regardless of the actual noise type. Our neural network-based SLAM algorithm is applied to a robot that has sonar sensors. The various experimental results demonstrate that the neural network-based SLAM guarantees a consistent environmental map under sonar readings that in general are known to have poor bearing accuracy and resolution.