Sign in:

Thousands of Tiny Robots

The Modular Robotics Blog

Hardware is Hard

We’ve been delaying the date that we think we’re going to ship our first Cubelets kits for months now, and it’s driving us crazy. It’s annoying other people as well, and I’m sorry for that.

I studied architecture in college, but have worked mostly on software projects since then. While software isn’t always easy, it can certainly be fast. Especially now that much software is internet-based, programmers can easily iterate, fix problems, and release new versions quickly. Most architecture projects are the opposite: five years from idea to final construction isn’t out of the ordinary. Although I originally had a thought that Cubelets manufacturing could be agile and more like software production, it’s clearly more like an architecture project, and in fact it’s been about five years since we started.

We received an enormous box full of thousands of tiny magnets yesterday, and they were all about 0.2mm too long. This seems like a tiny amount, but since our magnets are cast into the plastic Cubelet shells, magnets that are too long prevent the injection mold from fully closing and can also scratch the expensive mold badly. These custom magnets came after working with the supplier for three months to get the design just perfect. We’ll have to either shave down every magnet and re-plate it, or make a new batch which takes 2 weeks. It’s the little problems like this that multiply for Modular Robotics. There are 300 different parts in each Cubelet kit, many of them custom, from almost 100 different suppliers. And since assembly happens in sequential steps, each problem holds up the process. We’ve already sorted out problems like wheels that were too small, metal parts with lead in them, faulty LEDs, and spontaneously discontinued electronic parts. I never even imagined that something like an earthquake in Japan would influence our production. Although we’re successfully solving each new problem as it comes up, we’re going slower than we hoped.

We’ve finished building circuit boards for the first 100 Cubelets kits and the software is tested and ready. When we have all of our custom-made parts in hand and can begin mechanical assembly, I’ll be able to update this post with a revised shipping date. Meanwhile, thanks for joining us on this exciting little journey.

Join the Conversation  7 Comments
  • Carl

    At this point, ensuring you deliver a product that lives up to expectations is the important thing. More waiting for your customers followed by Cubelets they’re delighted with is infinitely better than them getting an already late product a bit quicker and then finding it doesn’t work as well as it could have.

    So don’t skip one little bit on the quality-control. It’s not worth it for your customers or your business.

    Continues to wait patiently…

  • Rikki

    If you open sourced the design (a la RepRap), you’d have a community of people fine tuning the the design and sourcing the parts.

  • We can wait…Rikki is on to something here as well…As an inventor, IT Manager & local business owner, I adhere to the saying ‘beware of advice that you pay for from strangers, but listen to advice freely given by family and friends’ The former is objective but solicited, the latter is biased but balanced by care and concern. We will be here waiting!

  • Hi,

    I have to agree, i do not mind waiting as long as they are perfect. If it is anything close to what I expect, there going to be great,

    Keep me posted on the progress

  • Saw your booth at Maker Faire this weekend in San Mateo and was impressed. Cool idea. I’m a research engineer at UC Berkeley in EE (wireless systems), and we frequently make prototypes or test beds but on a much smaller scale than what you’re trying to do. I empathize with the problems you’re having – I’ve seen all of them at least once. My pet peeve is the spontaneously discontinued electronics. Hardware development is indeed more of a challenge (and time sink) than might be expected. Hopefully things will smooth out when you’re into full production. Best of luck!

    By the way, do you have any papers on the original CMU research project that you’re willing to share, or can point me to?

  • Eric

    Thanks Fred. Here are a couple of papers to start, there are some more around that you might find by google-scholaring Cubelets’ old name: roBlocks.

  • Great, appreciate it. Right now I’m working on a project called Chiplets that’s part of a larger project for multicore rapid prototyping. We’re doing a test assembly that uses microbumped die on a silicon substrate (2.5D) but will eventually end up in a 3D structure, which we would have called Cubelets if you hadn’t gotten there first! Now we have to come up with a new name.