If the arm could automatically move based on what the camera is seeing, that sets the foundation for many interesting computer vision applications. Some will be very beneficial for competition.
If the drivetrain could automatically move based on camera input, then we could implement object avoidance and experiment with a combination of RGB object detection, the Realsense depth cameras and lidar.
This is where I got my inspiration for AI object detection: https://www.youtube.com/watch?v=JvAMI9OxQDU&t=85s
After watching that video, we may not even need lidar.