A modular hybrid localization approach for mobile robots combining local grid maps and natural landmarks

Abstract

This paper presents a hybrid localization approach for mobile robots combining local grid maps and natural landmarks. The approach at hand benefits from the advantages of both environment representations. While using memory-efficient geometric models describing natural landmarks as features for localization in structured regions, the proposed system clusters the remaining areas as raw local grid maps and incorporates those as pose features only for unstructured areas of the environment. To evaluate the functionality and performance of the approach at hand, extensive testing and benchmarking in an experimental setup has been conducted using an external sensor system for reference measurements

Mobile Robots for Moving-Floor Assembly Lines: Design, Evaluation, and Deployment

“…Specifically, we extend the feature-based Extended Kalman Filter (EKF) localization algorithm [14], which processes geometrical features extracted from raw laser scan data (e.g., lines and corners), to correct the robot's 2D pose estimate based on odometry information. In the automotive factory setting, it is also necessary to track the 2D poses of n dynamic objects (the cars to be assembled) simultaneously.…”