Algorithmic Approaches for Place Recognition in Featureless, Walled Environments.

摘要

This report gives a detailed presentation of three algorithms for automated place recognition by a mobile robot in featureless, walled environments where conventional feature descriptors are ineffective. The robot is assumed to have on it an inertial measurement unit (IMU) and a ranging device such as a light detection and ranging system. The environmental descriptors presented here are based on trajectory features and on junction features. The report shows how to derive the descriptor values at each point in the trajectory and how to compare descriptors recorded at different points in the trajectory to determine how closely they match. The output of the algorithms can be fed to an external mapping program and can be used to help correct errors in the map due to IMU drift. These algorithms can be combined with algorithms for fully featured environments to implement a nearly all-environment mapping system.