HUBO II on a stroll

Most humanoid robots tend to walk with bent knees. It makes them look like they’re trying to sneak up on someone at high speeds. HUBO II, however, has mastered straight leg walking which makes its ‘father’ Professor Jun Ho Oh very proud. Oh, leader of the Hubo Lab at the Korea Advanced Institute of Science and Technology (KAIST), recently gave a really great interview to Erico Guizzo of IEEE Spectrum. The professor explains the new walking technique, the robot’s dextrous hands, and its ability to upgrade. The HUBO II has an entire onboard computer available to run programs created by other developers. That’s an open invitation to faster innovation. Watch the proud papa with his mechanical son in the video below.

Along with other humanoid robots, like Honda’s ASIMO, HUBO represents an attempt to replicate realistic human movement and ability. To that end, straight walking is a great accomplishment. However, I think the possibility for incorporating third-party code (via the second onboard processor) may be the more important feature of the HUBO II. We’ve already seen how the HUBO was adapted to include Hanson Robotic’s realistic replica of Albert Einstein’s head. It’s possible we could see some similar, though perhaps strictly code based, cooperation from other teams in the near future. I’m certainly interested to see what the Hubo Lab comes up with next.

HUBO II starts to walk around 1:14 and shakes hands at 1:43. Hey, isn’t that the KAIST Star Wars walker in the background?

[screen capture and video credit: Erico Guizzo at IEEE Spectrum]
[source: Hubo Lab, KAIST]

There's a family resemblance, wouldn't you say?

It isn’t the first robotic chair we’ve seen, but at just over two meters tall and weighing 150 kg, the FX-1 is certainly the biggest. And the FX-1 just looks cooler than the OmniZero9 we saw at the 16th Robo-one competition, or the Panasonic mobile bed-chair. According to their website, KAIST hopes to use the FX-1 as a means of transporting the elderly, helping after accidents and disasters, and heavy lifting in an industry setting. The walker probably isn’t fast enough for military applications, but KAIST intends to speed up the robot’s pace. They also plan on giving the robot its own built-in battery, the ability to climb stairs, and increased carrying capacity – the robot can lift around 100 kg (220 lbs) for now. I’m not sure if the FX-1 will ever make it out of the lab, but it’s a great example of how man and machine may one day work together…to make nerds everywhere squee with delight.

[photo credit: KAIST, LucasFilms]
[video credit: KAIST, raingrove]

KAIST's HUBO (left) has many of the characteristics of Honda's ASIMO (right) but that's to be expected from two humanoid robots. Right?

Humanoid robotics is sort of like general artificial intelligence. There have been some amazing strides in narrow application robotics (such as Adept Quattro), and narrow AI (artificial writers), but the race to create human-like intelligence and bodies is still a long ways off. ASIMO and HUBO are two of the few high profile and well funded private programs actually chasing the dream of robots that can move and interact with people as an equal. Their successes, then, are typically unique in the field. ASIMO has had break-throughs in the level of complex human tasks it can accomplish. HUBO put on Albert Einstein’s face and interacted with people with help from Hanson Robotics. With enough time, effort, and money, it seems like that one of these two projects will eventually create a robot that can fully mimic human actions and find a role in human society. Companion bots may seem like just another science fiction fantasy, but these humanoid robotics programs are taking steps in that direction.

HUBO-2′s precise finger movements are pretty amazing. There’s a dexterity there that could make it a tough competitor to other robotic hands we’ve discussed. The smooth motion during the tai-chi segment is impressive for a research robot, but we’ve seen similar capabilities in industrial machines. It’s disappointing that the running segment was done without a complete upper body. However, I do give props to a robot that can swing a sword (1:51) even if it is made of foam.

KAIST’s robotic program is fairly young. Their KHR line of walking robots was launched in 2003 as little more than a pair of walking legs. HUBO-2, which is also known as KHR-4, shows how quickly the the line has advanced. It has 40 degrees of freedom (6 more than ASIMO), weighs just 45 kg (less than ASIMO) and can ran at speeds up to 3.3 km/h. ASIMO wins the last race hands down by running at 6+ km/h. It’s harder to compare their sensor and decision making capabilities. HUBO-2 has some remarkable voice and object recognition, with independent eye tracking. However, it hasn’t had a chance to test itself on a moving obstacle course the same way that ASIMO has.

Even if KAIST is actively copying Honda’s work with ASIMO (and I don’t think they are), that wouldn’t necessarily be a bad thing. ASIMO is a remarkable robot, we should be copying it, and trying to improve upon its design. Undoubtedly HUBO has many unique innovations that already differentiate it from other robots, but again, the uniqueness of the design isn’t the most important criteria. Success is. And I think that the more that robotics companies can borrow from, and compete with one another, the better the robots they build will be. Companies like Willow Garage advocate this explicitly by running open source robotics programs. Larger investors, like Honda, may not see the financial incentive in such open cooperation, but if it leads to the first truly human like robot, wouldn’t it be worth it? Sharing ideas and technology is a small price to pay to help create a Transformer. Or even a Gobot.

[photo credits: KAIST, Honda]
[video credit: PlasticPals]