It’s always exciting to open up a new shipment of Cubelet parts.  After the UPS truck has gone and we’re left with a pile of boxes, we tend to gather around with a bit of trepidation to see what’s inside.  Some of our suppliers are boring — they always send us what we order.  But our custom parts tend to be much more exciting.  We never know what will be inside an innocuous shipping crate.

Most of our custom parts are made at factories in mainland China.  We have a metal stamping factory in the North, a magnet factory in Ningbo, a PCB factory in Wuhan, and our injection molder in the South, in Guangdong Province.  Eleven other factories make things like gears, screws, spring pins, etc.  I’ve been working directly with our suppliers now for three years, and I’ll be the first to admit that coordination and communication is a serious challenge.  It’s hard to get the right parts.

Our main challenge now seems to be getting our suppliers to send us the same parts as they did for the last order.  Almost across the board, we’ll place a re-order (“please send us 10,000 more of the exact same widget as before”) and receive something different.  It’s weird: metal parts stamped from a different material, spring pins plated differently, PCBs of the wrong thickness, and plastic parts in the wrong color.  Honestly?  It’s driving me crazy.

The latest problem popped up on Friday soon after we received 12 boxes (twelve boxes!  it’s expensive to ship 12 boxes from China to the US!) of plastic Cubelet halves.  We began assembling them into Cubelets only to find that they were all failing our connectivity test.  The little test rig shown above is made of three Cubelet pieces on a sprung platform.  As we assemble each Cubelet, we pop it in here and a series of tiny LEDs lights up to check for connectivity and short circuits.  We’ve built a different jig for every Cubelet type that runs a specific set of tests on the electronics and hardware.

After some fiddling and measuring, we found that the plastic pieces were molded wrong.  The base thickness is supposed to be 5mm with a tolerance of 0.075mm.  And these were all 5.3mm thick.  0.3mm doesn’t sound like a lot, but it’s more than enough to prevent Cubelets from communicating with neighboring Cubelets.  The upshot?  All of the plastic casings are bad.  This is unfortunate for three reasons.

Resources.  It’s unfortunate because we now have a few huge boxes of plastic and metal parts that are now useless — it’s a shameful waste.  It took a fair amount of energy and petrochemicals to make these plastic bits, and now they sit.  We’re thinking about how we might go about cracking the plastic, removing the magnets, and recycling them separately, but even this process will require more energy input.  Normally, I’m OK with making our plastic widgets because I’m confident that they’re changing kids’ minds.  But making useless plastic widgets makes me feel guilty.

Money.  This mistake is also unfortunate because it’s expensive.  It turns out to be about a $40,000 mistake, and we don’t have piles of money on hand to deal with problems like this.  We’re funded by the National Science Foundation and a few other investors, and money wasted on stuff like this is money that gets diverted away from fun projects like new Cubelets.  The most irritating part of this for me?  We pay a Quality Assurance company to check every plastic part before it’s put in the box for shipment to us.  It seems that arrangement has not worked out.

Time.  The third effect of this problem is that it puts us behind schedule.  We have 250 outstanding pre-orders for Cubelets, eager buyers emailing every day, and the holiday season creeping up.  It actually takes about five weeks to get new plastic made: first, the magnets are custom milled, then they’re sent to the injection molder to be embedded into the plastic.  Then everything gets put on a plane to Boulder, CO.

Well.  I wanted to let you what we’re up to and why we don’t have a Buy It Now button next to the Cubelets on our site.  I hope it didn’t come across as a rant.  We’re adapting and getting replacement plastic and magnets molded and shipped as quickly as possible, and we’ll carry on making as many Cubelets as we can.  Hardware is hard.

In July, we shipped out our first production run of 100 “beta test” KT01 20-Cubelet kits.  That might not sound like a lot of robots, but 100 kits means 2000 Cubelets.  Each Cubelet has 6 circuit boards and 24 tiny magnets in it, so the quantities were actually pretty large.

Our first 100 customers were extremely patient.  Some of them pre-ordered Cubelets back in October 2010 and waited patiently while we dealt with some manufacturing, material, and supplier challenges.  But Cubelets are finally out in the world.

We’re busy manufacturing our second run (5000 Cubelets this time), but our early-adopter friends have already started posting reviews and videos of their first experiments and play sessions.  They’re definitely worth looking at.

• Jonas at Cornell posted a video of some creative robots he’s built.
• Carl in New Zealand writes a thoughtful review of the Cubelets KT01 kit.
• Per at the Flexibility Envelope shot an unboxing video with some initial impressions.

I gave a talk at the University of Colorado’s Computer Science Colloquium a couple of months ago. It’s kind of long (and totally unedited), but if you’re curious about the history of Cubelets and some other related projects, here you go:

I’m excited, and I’ll admit: a little bit amazed. Today we’re shipping out the first Cubelets kits to our eager pre-order community. I’ve been working on Cubelets since 2006; first as a PhD project, then as a fledgling idea in my garage, and now as a real company with a team of 9 people. Five years in the making, and today we’re carrying boxes down to UPS and watching them roll down the conveyor belt. It’s been so long in the making that I keep pinching myself to make sure I’m not dreaming. We’re really shipping out products!

Assembly has taken us a lot longer than I had estimated (even though experience in these things has taught me to triple any time estimates). We were pretty accurate in planning for all the “real” assembly tasks like soldering, screwing, folding, calibrating, and programming, but all of the other miscellaneous tasks crept up on us: things like opening boxes, organizing parts, carrying things around, building shelves and charging batteries. Yesterday I saw something that none of us has ever seen before: we killed an entire one pound roll of solder.

Today we’re going to happily ship out as many Cubelets as we can. But we know that on Monday, things will change. Customers will begin emailing with technical questions, the phone will ring, and our little inward-focused design team will begin looking outward; dealing more with educators, kids, toy stores, and customers. This will no doubt be a transition with its share of hiccups, but if we can finally get Cubelets out the door, we can do anything. Today, we celebrate!

Maker Faire was something. We were expecting a crowd, but we weren’t expecting the madness created by thousands of kids clamoring for Cubelets over the two full days. It was amazing! Although the four of us that went to San Mateo are still a little exhausted, I’m happy to report that Cubelets are indeed tremendously fun to play with. Everyone was excited to build robots, from a few four-year-olds to a huge mass of 7-13 year olds, to a bunch of nerdy adults, and not just a few robot scientists. The mass of kids literally pushed our table back — we had to slide it forward every hour or so. And we got three Education Awards! I was so busy demo’ing that I barely had time to walk around, but there are a few photos of kids and robots on our Facebook page.

We learned a lot at our first Maker Faire and are looking forward to coming back with a bigger, badder (giant inflatable Cubelets?) setup. In the meantime, I’ve updated our FAQ with some of the more common questions from the last weekend. See you in NYC in September?

Joy. After months of delays and unexpected problems, we’re finally building Cubelets. In the photo above, Brandon is snapping circuit boards into the plastic casings and testing each face. It’s surreal watching finished Cubelets come off our assembly line — I almost can’t believe it’s really happening.

If you’ve pre-ordered a Cubelets Standard Kit, we’ll ship you your Cubelets on June 17, 2011. Thanks for sticking with us!

If you’d like to get in on our next manufacturing run of Cubelets, modrobotics.com will be open again for pre-ordering on June 1.

It’s been a lot of fun watching submissions come in for our Cubelets Packaging Design Competition. We received 12 entries, most via email, although a couple of the submissions were physical, and one was a video entry. Sadly, no interpretive dance entries. We had a great time judging and discussing the relative merits of the different approaches and styles.

I’m happy to announce that we picked Helle Schou Pedersen’s design (above) as the winner. It’s a simple design — a cardboard box with labeling and some cutouts — but it is elegant, sustainable, robust, and inexpensive to produce. Helle sent her entry from Denmark and she’s received a prize of US$1000 and a Standard Kit of Cubelets.

Second place goes to Liam Ward, whose design (above) uses a molded paper pulp inner box with a printed paper sleeve. Molds like this can be made from recycled material, although we haven’t found a supplier that’s interested in making any less than 20,000 boxes of one type. Liam’s design might not be easy to use in practice as the strong magnets would click the Cubelets together, preventing them from sitting nicely in their little holes, but the dividers could easily be removed to allow a solid 20 Cubelet “brick” to nest in the paper pulp. Liam sent his entry from the UK and receives a Cubelets Standard Kit.

Here are some photos of a few more outstanding designs. Thanks for participating!

We’ve had our first set of injection molded Cubelets in the office for a number of weeks now. We’ve been putting them through their paces in an attempt to identify and rectify any issues in the design or hardware of Cubelets. In an attempt to simulate the regular use and abuse of our product we have performed numerous tests on the blocks.

One such test is impact testing of Cubelets. Government standards require that we do impact testing from a specified height onto a highly specified surface. We must drop our product from 36 inches onto â…› inch type IV vinyl tile (the type you would typically find in schools). I set up a nice drop testing studio in our stairwell here at Modular Robotics with the specified tile and some nice lighting.

We performed 4 successive drops of our blocks, each time checking for damage to the cube. After each drop of the block we ensured that no sharp points or broken magnets were created during the impact. We also tested the electrical and data connection of the six faces of the block after the testing. These impacts had no adverse effects on the block’s performance. We were pleased to show our expectations of Cubelets to be able to withstand this type of fall. Below is a set of images from one of the drop test videos taken during the testing. We continue to do testing on different aspects of Cubelets including impact tests on our final Battery block design. Keep checking back as we will keep you up to date as we test different blocks in different ways such as fatigue or strength tests.

We had a little party Thursday afternoon for our local Boulder friends. Happily for us, some brought kids. Amidst a crowded hubbub of geeky grownups Elliott Dobbs (he’s six, with one more month in kindergarten) and Arthur Correll (he’s five and a half) sat there building robots! Right now we have only one Cubelets kit–the all-black engineering prototype that’s in our video–so they had to share, but that worked fine. At first they just snapped blocks together randomly to see what would happen. But pretty soon they began to understand the different functions of each Cubelet and the gradient dataflow model that makes the robots go. With single-minded focus on building robots they worked quietly for over an hour and a half (ignoring the Lego Mindstorms robot and other toys on the table next to them). And for me that was the best part of the party–to see young kids playing with Cubelets entirely entranced.