How do robotic mice navigate to solve mazes in the shortest time? Three interesting strategies invented in the Micromouse competition history.
How do fully autonomous robots solve a maze efficiently? Veritasium introduces the Micromouse competition, a maze-solving competition where robotic mice compete to see which one navigates and solves the maze in the shortest time.
In the competition, each mouse can make five attempts at reaching the goal from the starting point within a time limit. The common strategy is using early runs to search and map out the maze, aiming to find an optimal route from the start to the goal. In subsequent runs, the mouse traverses that route in the shortest achievable time. The most common search method is the flood-fill algorithm.
There are many remarkable strategies invented throughout competition history. Here are three I found the most interesting:
- The shortest path may not be the fastest. The initial instinct to solve a maze fast is to find the shortest route. However, there are instances where the micro-mice actually took longer to reach the goal using the shortest path. This is because turns slow down the mouse, and a longer route can require fewer turns, making it faster.
- The turning mechanism employed by robots is critical. Previously, all micro-mice used to make 90-degree turns when maneuvering through the maze. However, an innovation introduced by mitee3 involved turning at 45-degree angles and traveling diagonally. This method proved to be much faster because it reduced the path length and introduced more possible routes.
- Equipping a vacuum fan under the mouse can greatly enhance its performance. In physics, centripetal force is necessary to keep you on track when driving fast and making turns. The faster you move, the more force is needed to maintain your trajectory. For micro-mice, incorporating a suction fan can provide additional centripetal force, enabling them to make turns without significant deceleration. Compared to a non-fan mouse, the newer one can achieve centripetal accelerations of 6g or even higher.
It can be quite nerdy, but I highly recommend watching the whole video. It perfectly demonstrates that robotics is not solely a software or hardware problem. It's both.