One game, many robot categories
RoboCup uses soccer as a shared benchmark problem for advancing AI and robotics, on the idea that a game everyone understands makes a demanding technical challenge easy to follow and easy to judge. Robots have to perceive a moving ball, other players, and field boundaries, coordinate with teammates without a human calling the plays, and make split-second decisions, all while walking, rolling, or otherwise staying upright and mobile. That combination of perception, real-time decision-making, and physical control is exactly what makes it such a demanding testbed for general robotics research, not just a novelty.
The competition spans several leagues that use very different hardware, from small wheeled robots playing on a tabletop-sized field to full-size humanoid robots attempting to walk, kick, and recover from falls, plus a simulation league where teams compete entirely in software. Each league isolates different research problems, so a small-robot team might focus heavily on multi-agent strategy while a humanoid team spends most of its effort just keeping the robot from falling over.