Position Ground Truth For a Round Robot Using LiDARs

Localization is an essential role a robot must have, as it must be autonomous and require less human intervention when operating in society. When developing and testing a localization system, it’s crucial to compare what it calculates with a ground truth provided by a sensor not being used in the system developed. Thus, this paper proposes a ground truth approach using two LiDARs to detect with reasonable accuracy the actual position of round robots in the real world, like the ones used in the Small Size League of RoboCup. The league is a considerable scenario for tests since it uses a particular camera system positioned above the field to detect the robot’s position, and the internal sensors, like encoders, cannot provide high accuracy due to wheel slipage. The software architecture includes a scan matching step to combine LiDARs, and the detection step is made through the Hough transform to recognize circular patterns in the image created with the LiDAR points. The results showed that the system based on the LiDARs is significant as ground truth, with a mean error of less than 1 centimeter in the 17 positions analyzed in tests and a standard deviation of around 1 centimeter. The results are notable when compared with the league’s camera system in these scenarios. The system can be adapted to different sizes of robots and be used in any indoor environment, although the Small Size League was the main scenario for the tests.