We were very excited to receive our acceptance email this morning, as we’d already planned a team meeting (at 10yo’s insistence) for that time. So we had our discussion.
Discussion point 1: Chassis design
Glitterator mark 1 was LEGO-based, and when we put in the application for this year’s entry, we weren’t sure if we would continue this theme. The biggest problem with a LEGO-based robot is that LEGO is designed to be disassembled and reassembled over and over again, which is great for model builders, but has the unfortunately side issue that sometimes it becomes disassembled when you least want it to be… like right in the middle of Pi Noon!
On the other hand, a LEGO-based design is easily adaptable, and our experience with LEGO-based pi-controlled robots didn’t end with Pi Wars 2017: Peter and Angus have spent the summer building a series of pi zero controlled LEGO robots. Some with tracks, some with wheels, and even some with legs. The legged ones were fun, but not really practical (slow and difficult to manoeuvre), but the tracked and wheeled ones have been very effective. (Although the two-wheeled version was given up as too difficult.)
So in the end, we have decided to go with our experience, and use a LEGO-based design again. The big difference is that we will use Technics LEGO, which is considerably more difficult to disassemble, and so hopefully will be more robust. Related question was tracks or wheels? I was quite keen to try wheels, with servo-based steering, but in the end was overruled, so we’ll have tracks again. Different type of LEGO tracks last year – rubber ones last year – and I’m concerned that what we have now are not quite grippy enough, but the plan is to add some rubberised tread to the links, so I think it will be OK.
There are some other particular aspects of chassis design that we went on to discuss too… but we have to keep some secrets 😜
Discussion point 2: Control
Last year, we used a genuine PS3 controller. It’s a bit fiddly to get paired, but once we’d figured that out, it was working pretty well, right up until the day of the competition, when it kept conking out! Turns out it wasn’t just us, but lots of people had similar problems, and not just with PS3 controllers either. In any case, with the explosion of robots in the house over summer, we’ve needed more controllers, and the PS3 ones are just too expensive. Instead we’ve been using Rock Candy controllers, which again, have been working fine. But some people reported similar issues to ours with Rock Candy controllers at the venue.
Sooooo… there’s been quite a bit of discussion in the community about what causes the problem. The most plausible explanation that I’ve come across, that I haven’t put to the test, is that our power supplies are not perfect, and that sometimes the voltage drops a bit below 5V – not enough to stop the pi itself running, but enough to cause glitches in the peripherals. So this is something we will take into account. We’ll also be looking at alternative controllers, but I’m not so sure this will solve the problem, especially if the power glitch hypothesis proves to be the cause.
Discussion point 3: Raspberry piS
So last year we had a pi 3 doing most of the work (motor control, sensing, etc), communicating (via USB) with a separate pi zero that controlled our light show. This year, one thing that I did put in our application was that we wanted to explore this theme further. Now, over summer we’ve developed a very nice base for pi zeroes with remote control for LEGO robots, and I think this will work very well as the base of Glitterator mark 2. But I’d like a separate pi – probably a pi 3 – to do the autonomous control, sending low level movement instructions to the underlying pi zero. And again we plan to glitter, so another pi zero for light control. And then the duck shoot – separate pi zero as well? Well, this discussion didn’t reach firm conclusions about the number of pis, but a separate one for autonomous control seems a fairly firm plan right now.
So, this has been a rather brief summary of what was a rather long (and sometimes heated) discussion. But the result of it is an action plan.
- I will be experimenting with various sensors – some we used last year, some new – to decide which we will be using for various challenges.
- Angus will be assembling a Technics-based chassis, with ground clearance, and some secret features that we want to test.
- Peter will be looking at what we can do with the camera. He’d like to use it for more than just the rainbow challenge; I think other sensors will be more effective. But we’ll experiment and see.
- And Erin, who is just 7, has declared that she wants to help with the coding this year. So she’s starting to learn Python! (Angus did it at age 9.) So there’s a challenge that she and I will work together on (with the help of the brilliant Kano tutorials). (She did actually get a taste of Python at a Pi Jam workshop run by Ben Nuttall last year.)