3DoF AR Tag Following Robot

Our team engineered a sophisticated 3 Degree of Freedom AR tag tracking robot, featuring a custom-made chassis crafted from 80/20 T-slot aluminum. Every aspect of the robot’s design and construction is meticulously tailored, showcasing a commitment to precision and innovation.

The primary objective of this project was to demonstrate the distinctions between position-based visual servoing and image-based visual servoing. The robot incorporated an ELP 8mp USB camera seamlessly interfaced with a custom Python program. Calibration of the camera was performed utilizing OpenCV’s integrated calibration function and a dataset comprising 20 sample images of a 9x6 checkerboard.

In the realm of position-based visual servoing, OpenCV transmitted the AR tag coordinates directly to the movement algorithm. The robot, responding with remarkable accuracy, would adjust its position by 30% of the distance towards the target position above the tag. While effective, this method demanded significant computational resources.

Conversely, in the domain of image-based visual servoing, OpenCV identified the four corners of the AR tag. These corner coordinates were then fed into the movement algorithm, leveraging a timestep of 1 and a Jacobian matrix to orchestrate the robot’s precise movement to the desired position. This approach, though faster, exhibited a tendency to overshoot and faced challenges when confronted with substantial changes in the AR tag’s position.