Mapping While Following: 2D LiDAR SLAM in Indoor Dynamic Environments with a Person Tracker

TL;DR

This paper introduces a framework for 2D LiDAR SLAM in indoor dynamic environments that effectively tracks and filters out moving people to improve map accuracy without manual control.

Contribution

The paper presents a novel framework combining people tracking, filtering, and following to enhance 2D LiDAR SLAM in dynamic indoor settings, reducing mapping errors.

Findings

Effective exclusion of dynamic objects improves map quality

Framework reduces mapping errors in dynamic environments

Verified with two classic SLAM algorithms

Abstract

2D LiDAR SLAM (Simultaneous Localization and Mapping) is widely used in indoor environments due to its stability and flexibility. However, its mapping procedure is usually operated by a joystick in static environments, while indoor environments often are dynamic with moving objects such as people. The generated map with noisy points due to the dynamic objects is usually incomplete and distorted. To address this problem, we propose a framework of 2D-LiDAR-based SLAM without manual control that effectively excludes dynamic objects (people) and simplify the process for a robot to map an environment. The framework, which includes three parts: people tracking, filtering and following. We verify our proposed framework in experiments with two classic 2D-LiDAR-based SLAM algorithms in indoor environments. The results show that this framework is effective in handling dynamic objects and reducing the mapping error.