Скачать или смотреть Autonomous Calibration Challenge - Arena rUNSWift

https://spl.robocup.org/rc2021/

00:00:00 Setup
01:09:56 Berlin United - NaoTH
02:07:00 HTWK Robots

This challenge deals with the problem that every robot and every field have small differences, even though all robots in our league are the same model and the fields are built to the same rules. For example, there are differences in the environmental conditions such as lighting, and robots have “wear-and-tear”, or different manufacturing tolerances. These differences greatly affect the robot’s ability to play soccer.

Careful calibration is required to adapt every individual robot to these small differences. A badly calibrated robot will fall over when it tries to walk, and won’t be able to see the ball. Calibration works out ideal camera settings or the exact “offsets” of the manufacturing differences, and impacts of joint wear. Usually calibration is done manually by educated guess-work, or a semi-automated learning procedure that requires significant human assistance. It is far better and quicker if the robot could “automatically” calibrate itself. Thus, this challenge requires the robot to “self calibrate”.

The challenge consists of two phases: In the first phase, the robot has the opportunity to collect data from which it “self calibrates”. In the second phase, the robot performs a series of tasks to judge the accuracy of the calibration. These tasks are walking to fixed positions and stating the positions of two soccer balls that are placed on the field. The robot is judged on speed and accuracy, as it must: both walk stably and accurately perceive its environment as quickly as possible.