Pi Wars 2020 or Bust!

27th May 2021

Now that the new theme and challenges have been announced, The Tauradigm team have been reviewing them against our previous build to see how much will need changing or additional attachments needing building. The challenges this year are:

  • Lava Palaver
  • Pi Noon – The Hindenburg Disaster
  • The Zombie Apocalypse
  • The Obstacle Course
  • Eco-Disaster
  • Escape Route
  • Minesweeper

I will do separate blog posts on how we’re hoping to approach each challenge, but briefly:

Lava Palaver
The straightish line speed challenge from last year, near with a vertical speed bump too. We think the sensors fusion of IMU, distance sensors and vision will still work to give a fast time. We may need to detect for wheelslip or taking off in case that ruins the control algorithm

Pi Noon – The Hindenburg Disaster
Again, the same balloon popping event of the past few years. Without knowing what the potential spiky obstacle is, our sensor array should be sufficient to detect the walls and the opponent.

The Zombie Apocalypse
This new challenge is a modification on the popular duckshoot and space invaders challenges, where targets need to be hit with soft projectiles. Our autonomous aiming showed great potential two years ago, and the new flywheel launcher is looking good, so we’re hopeful on this one. We need to add a tilt mechanism as the targets are no longer at ground level. It’s not clear from the rules what the targets will be, or if they’ll be moving, both of which could give us additional software challenges.

The Obstacle Course
no new information on this challenge so its hard to tell what we’re in for. The ‘travellator’ obstacle last year would have been quite tricky to navigate autonomously with the approach we had planned. Now we know it may feature, we can probably incorporate a special case to deal with it.

An interesting new challenge where coloured barrels need to be collected and moved to designated zones. We’ll need to build a new attachment to deal with lifting/moving the barrels, which might be quite complicated. This is also looking like a difficult challenge to code for, as there’s shape and colour detection required, coordinated moves and path planning needed. Should be interesting!

Escape Route
A variation on the minimal maze and canyons of mars, this is a maze like challenge with no prior knowledge and probably not continuous walls. We had planned to preprogram both the shape of the maze and the target course, then use the sensors and a localisation routine to figure out where we were in it and how to keep to that course. We’ll now have to use SLAM (simultaneous location and mapping) to both work out the maze shape and where we are in it, as well as have a path planning routine. in some ways this challenge is quite similar to the Eco-disaster, there could be a lot of shared code, certainly the learning on the tricky parts will help.

Another new challenge, a little like a BATAK test used for human reaction times, where the robot has to drive to areas lit up. We could use the camera to detect the lit up panel and so be able to attempt this challenge with no new hardware, but we think it will be much more reliable and probably faster to use a dedicated light sensor attachment.

On the hardware side, we’ve been making good progress on the pcb soldering and are hopeful we can start a drive test and systems check in the next week.