Benchmarking the Localization Accuracy of 2D SLAM Algorithms on Mobile Robotic Platforms 2020-01-1021
The information regarding the position of a robotic platform relative to its environment is essential for a multitude of subsequent applications such as obstacle avoidance, path planning, navigation and motion-control. In recent times, numerous Simultaneous Localization and Mapping (SLAM) algorithms have emerged, including easily deployable ROS implementations. These solutions differ in their implementation techniques and often depend heavily on the quality of sensor housed on the robot. The objective of this paper is to help users make an informed decision on the SLAM algorithm to use depending on the type of hardware available and the desired final application. We analyzed four different SLAM algorithms that are currently deployed in ROS and are extensively used in various practices in robotics: Gmapping, Hector, Karto and Cartographer. The accuracy with which the four SLAM algorithms can localize a differential drive robot in a controlled indoor environment was benchmarked against the OptiTrack motion tracking system. The OptiTrack motion capturing system, using Prime13 cameras, is a powerful 3D motion tracking tool that is capable of finding the pose of an object in 3D space with millimeter level accuracy and precision. Finally, a comparison of the performance of these SLAM solutions has been provided. Our focus is on the accuracy of robot trajectory-following as generated by the SLAM algorithms.
Mugdha Basu Thakur, Matthias Schmid, Venkat N Krovi