HUBOway

After a few days here, I found out that Dong-il Choi (ME Masters student) has been working on the so-called "HUBOway" for the past 1.5 years as a thesis. The project is meant to be a simplified, COTS version of the segway. After 6 months of development time, he built the first model; since then he has made 6 more, some of which are used at other universities.

As it turns out, Dong-il has accomplished what our team could only dream of: full balancing and command input with optimal control. It uses Pb-Acid batteries (24V), S28-150 motors, a TI F2808 DSP, and 2 OSMC's. It's about half the weight of our version, uses only 2 belts /motor for speed reduction, and requires a fraction of the machining time to build. Those rumors about these guys being the best of the best are so, so true. The contrast between HUBO and DASL has never been more shockingly obvious.

Besides having the best people in the country, there seems to be several main reasons why this lab is so advanced:

1. Parts. The sheer volume of electronics, connectors, SMD components, nuts, bolts, springs, and everything else is staggering. There is almost an entire room, at least 12'x12', devoted to wall racks and bins full of components. As I've said before, if you want to build something rapidly, you can't wait around for several days every time you need some trivial piece.
2. Standard Electronics. As part of the HUBO project, the lab has developed a simple, effective 48V motor driver based on the F2808 DSP. These motor drivers are used for almost all servomotor applications, including other robots, experimental legs, and small projects. Similarly, they have a standard 2808 DSP breakout board that is used for all of the small robots. This chip is powerful enough for almost any small robot, and the board is only 1.5"x2". It has a standard plug, and so any module can be plugged into any board, making it simple and easy to replace or upgrade. Compare this to the PICs that our lab uses, which have caused untold weeks of delay due to lack of computing power and lack of peripherals (see Vefa's project).
3. Machine Tools: The CNC machines available to the lab are capable of producing robot parts quickly and accurately. Given the time usually spent on solid modeling, being able to tell a CNC machine to simply make a part as specified would save hours of drawing and manual machining. If multiple identical parts need to be made, all you need is more stock.
   
4. Advising. Dr Jun Ho Oh seems to be around almost constantly, giving students technical advice and helping to solve problems. The PhD students also help advise the master's students. This way problems are solved quickly, before too much money and time is spent barking up the wrong tree.

The price of all this infrastructure cannot be overlooked, as likely $10,000,000+ have been spent on tools and equipment. Regardless, what this lab demonstrates is just how far you can go if students' time is not wasted solving problems that shouldn't exist in the first place (such as lack of computing power), or problems that have already been solved (such as how to machine a part accurately).