Evaluation of SLAM algorithms in realistic sensor test conditions

Fang, L 2018, Evaluation of SLAM algorithms in realistic sensor test conditions, Masters by Research, Enginnering, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Fang.pdf Thesis application/pdf 19.89MB
Title Evaluation of SLAM algorithms in realistic sensor test conditions
Author(s) Fang, L
Year 2018
Abstract Autonomous robotic systems rely on Simultaneous Localisation and Mapping (SLAM) algorithms that use ranging or other sensory data as input to create a map of the environment. Numerous algorithms have been developed and demonstrated, many of which utilise data from high precision ranging instruments. Small Unmanned Aircraft Systems (UAS) have significant restrictions on the size and weight of sensors they can carry, and light-weight ranging sensors tend to be subject to greater error than their larger counterparts. The effect of these errors on the mapping capabilities of SLAM algorithms will depend on the combination of algorithm and sensor. To quantitatively determine the quality of the map, a map quality metric is needed. This thesis presents an evaluation of the mapping performance of a variety of SLAM algorithms that are freely available in the Robot Operating System (ROS), in conjunction with ranging data from various ranging sensors suitable for use onboard small UAS. To compare the quality of the generated maps, an existing metric was initially employed, however deficiencies noted in this metric led to the development of two new metrics. A discussion of both the existing and new map quality metrics, and the advantages and disadvantages of each, is presented as part of this thesis. To evaluate the performance of algorithm/sensor combinations, ranging data was collected from various sensors in a known environment. Both sensor poses and the ground truth map were obtained using a highly-accurate motion capture system. The measured sensor poses were then corrupted with noise and drift to simulate odometry measurements required for the SLAM algorithms. Of the SLAM algorithms tested, Gmapping was found to produce high quality maps with wide-field-of-regard range sensors in the presence of odometry noise and drift. KartoSLAM produced similar maps to Gmapping (with wide field of regard sensors), though it did not cope as well with odometry errors. Hector Mapping tends to excel at creating maps with wide field of regard ranging sensors.
Degree Masters by Research
Institution RMIT University
School, Department or Centre Enginnering
Subjects Aerospace Engineering not elsewhere classified
Control Systems, Robotics and Automation
Photodetectors, Optical Sensors and Solar Cells
Keyword(s) SLAM
Map quality
Version Filter Type
Access Statistics: 77 Abstract Views, 317 File Downloads  -  Detailed Statistics
Created: Mon, 03 Dec 2018, 13:03:29 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us