Autonomouse’s automatic function (if it worked) seemed cool to replicate in the antweight division, but to make it more reliable (and safer), a manual mode would need to be added. Of course, the robot would only be a test pusher to start with.
Weight would be the issue. Other than that, I have no clue if there are any AI engineers in the forum.
Since autonomous robots aren’t covered in the AWS rules it’s something you’d have to ask event organisers about. From my perspective, as the operator of a silly little sportsman antweight arena, I would want to see clear demonstration that the robot failsafes into manual mode and stops moving when you switch the transmitter off, just like any other robot. I would think most EOs would agree with that.
Beyond that, I don’t see an issue with autonomous robots in the arena. The hard part is making it actually work. Maybe start with something relatively easy, like an infra-red or ultrasonic proximity sensor that triggers a flipper or an axe? That’s a pretty good beginner Arduino project - it’s only about a tenner in parts and a few dozen lines of code.
1 Like
An automatic grabber would be deadly (in a robot combat way, not actually deadly to people.)
I was thinking about a robot that would have 3 distance sensors to prevent you from driving into the pit accidently but when you start researching it gets complicated very quickly. Main issue is the weigh as someone just mentioned, you would need some more capable microcontroller and unless you design your own PCB, buying ready products to handle the voltage change from your battery, motor drivers and the sensors very quickly starts adding up.
That means that whatever your automation is, it should be worth more than that weight in armour/weapon. This is where another challenge comes in - making it reliable is a big task, when you start reading about it, each sensor type has it’s limitations. Especially used in an environment when you constantly crash into something. You would have to take all that into account plus add anything else you can predict (in my case for example a distinction between detecting the floor and robot simply being upside down) and compensate for it in code, which is nights worth of work probably.
If you’re thinking AI then it gets even more complicated in terms of weight. Even the simplest computer vision would mean that if you’re lucky your robot would be a Raspberry Pi with attached wheels plus a battery and that’s it - no room for armour or weapon :). You could probably have some capable device outside the arena with a camera that does the hard part and then just sends it back to the robot, but that has it’s own problems. In either case the bigger part would training it (even assuming it’s allowed).
To even open Isaak Lab or similar software, you would need a computer that with current RAM and GPU prices would be hard to justify buying just to potentially win a beer coaster :). This might change in the future but at the moment my opinion is that for antweight class it’s better to focus on mechanical engineering than going toward autonomous actions. I’m curios to see what the future brings and if it will become a standard soon but at the moment I would say that if you’re interested in automation/autonomy separate the two, focus on the design of your combat robot and build a separate one like SO ARM 101 to learn and experiment.
1 Like