Compare Open Source vs. Proprietary Localization Software | Tech News
One of the most common things to do with a new robot is enable autonomy, or at least get started on it with Simultaneous Localization and Mapping (SLAM). For years, the easiest way to do this in ROS has been to utilize the open source Gmapping package, a ROS wrapper for the Gmapping software created by OpenSLAM. Gmapping provides an easy and relatively straightforward way to begin mapping and localizing your robot.
However, just because something is free (or open source), doesn’t always make it better. With experience using gmapping ourselves, and hearing so many of our own customers and the greater ROS community lament the inaccuracy of gmapping for their projects, we knew there should be an alternative.
In response, last year Clearpath introduced the Autonomous Research Kit (ARK) to the robotic research market. ARK is a software and hardware package that utilizes the industrial grade autonomy software developed by Clearpath’s material handling division, OTTO Motors. The OTTO Motors team have put hundreds of thousands of hours into developing, testing and optimizing, and the ARK is a research enabled, black-box version of that software.
While ARK has been available on Clearpath robots for several months, our R&D team wanted to pit it head to head with gmapping to determine just how our proprietary localization algorithms stacked up against the open source incumbent.
Using a Husky UGV with dual SICK LMS111 lidar, several tests were run with both sets of software, including manual and autonomous operation, after setting up a VICON motion capture system as a ground truth.
By analyzing the resulting datasets, we are able to see how each software localizes the robot compared to the motion capture data. From this, we can determine which has the lowest error rates, as well as see which one provides the most amount of data points.
For the full report with testing results, as well as access to the collected ROS data, click the button below.