Inspiration/innovation in weaponry

[adamclark]

Now heavily into a 1kg unit that has 6 radial sonars driven by an arduino, compass, gyro and accelerometer, all feeding into a wifi raspberry pi with I2C on a lipo pack. Driving two 25mm motors with the ubiquitous Chinese eBay controller on 70mm skateboard wheels. 1kg to spare for spikey bodywork and should have the first autonomous hunter ready. Target is 2kg for a deploy and first units looking like a dinner plate spikey mine. Oh and they have fricking lasers as well.

Remote vision and recognition with the pi cam and openccv is well documented as an option.

Thinking here is that with 3 robotic judges and an educational remit from the Beeb, would I get a a selection for the next series. As a TV producer would I pick a big fat hardox remote control box driven by a nerd, or a complex autonomous machine overseen by a debonair swashbuckling raconteur controller of a fleet of mad robots

[Giles]

What is your plan for pit detection and tracking other bots? I did a bit of thinking about this not so long ago and was leaning towards something like RANSAC with lidar, using outliers in the data to determine where other bots are while building up a 2D map of the arena. That would probably be quick enough, considering finding the outliers is a direct by-product of running ransac to detect walls/corners. Sonar is also quite short range and has a broad angle of detection, have you considered how this might effect the bot?

[Pinski1]

If you can crack using SLAM to build up very high update rate pictures of your surroundings I think there is a couple of million/billion dollars in it for you. Check out the robotics DARPA grand challenges to get an idea of the difficulties.

[TechAUmNu]

The fact there are two house robots will also throw it off big time. How is it supposed to work out which robot to attack? Unless your lidar is high up and really expensive it will never see the pit, so the only way to get round that would be to know the layout of the arena and just figure out where you are so you never go near it. Maybe when we have small solid state lidar it will be more feasible. Something like this but much smaller http://spectrum.ieee.org/cars-that-think/transportation/sensors/quanergy-solid-state-lidar

I think the most that autonomy can do at the moment is either automatic turning like Chomp from BB or avoidance of robots ramming into you, it would only work if you have something like Omni wheels so can get out of the way without turning. Which could be reasonably effective against spinners.

[adamclark]

Well AFAIK the pit is always in the same position, as are the walls. Starting positions are also fixed. So once you know where the wall is and roughly where your starting square is then you are pretty much set. House robots will tend to be in areas that are also generally known so they can be downgraded areas to enter. In terms of tracking other robots, if they are not emitting ultrasonic (which i'm working out how to listen instead of ping in between scans) and are within 2m then advance,otherwise just scoot about. Lidar modules seem a bit too expensive, although there is a cheap laser/webcam parallax unit. However the sonar units are only £1 each, so cost trade off is a big benefit

[daveimi]

It's a brilliant idea. I think event organisers will take issue with failsafing and weight.

[adamclark]

I think I've been hypnotised

[Giles]

So you will be going with a "randomly drive around until you see something" kind of approach as opposed to landmark/object tracking? That will work with ultrasound no problem but I wonder what hilarious situations this will cause

[killercarrot]

... a debonair swashbuckling raconteur controller of a fleet of mad robots

+1 for Jumbo Johnson

[adamclark]

So you will be going with a "randomly drive around until you see something" kind of approach as opposed to landmark/object tracking? That will work with ultrasound no problem but I wonder what hilarious situations this will cause

I think I might try that, a bit of patrolling. Also have a couple of PIR that seem to be quite efficient. Nice video on SLAM and odometry mapping , https://www.youtube.com/watch?v=wVsfCnyt5jA to digest. I think with such a finite area it might well be possible to determine where we are to a certain degree. I'm going to play with the IMU this weekend, but I think with a hunk of steel it probably wont be of too much use. I remember I had problems before when I had one and a video link and gyro inside corporal punishment. If the BLE chipsets were not so expensive then a shared map array would be nice (call it cloud based to be on trend)

FWIW, an in progress sonar module , single channel, showing 5cm range (20 scans per second) and a PIR trigger of YES (the camera waving above))
IMG_0891.jpg