WHEN Sony last month unveiled Qrio, a diminutive two-legged robot that can run like a human, it seemed like yet another breakthrough in a sequence of such advances in robotics. Only last year, a troupe of similar Sony biped robots performed a choreographed dance routine. And two years before that, Honda showed off a humanoid biped that walked with a confident swagger.
But while it is highly entertaining to see these robots doing their stuff, most robotics experts say Sony and Honda鈥檚 technology is little more than a red herring. They say walking and running are probably the least efficient, most power-draining ways for robots to get around 鈥� especially in the home. As a result, it鈥檚 going to take decades before we see humanoid biped robots traipsing around our houses carrying piles of washing or putting up shelves.
Hollywood and sci-fi authors have created a distorted image of what a useful robot should look like. Think C3PO and the Terminator. When asked to picture a robot, most people immediately describe a humanoid robot, says Ben Krupp of MIT spin-off Yobotics in Cincinatti, Ohio. 鈥淚t鈥檚 a dream. We want to see something walking around the house that looks like us,鈥� he says.
Advertisement
So why are blue-chip corporations like Sony and Honda, with interests in audio-visual consumer electronics and vehicle manufacturing respectively, investing so many tens of millions of yen in furthering biped robotics, a seemingly irrelevant field?
Sony describes its robots as 鈥渃orporate ambassadors鈥�. Meanwhile, Honda says its aim is to harness technology 鈥渢o improve the lives鈥� of its customers. But critics say the truth is that they have been seduced by science fiction.
鈥淭hey鈥檙e doing this because it鈥檚 fun. Sony is in the entertainment business,鈥� says Colin Angle, a founder of iRobot Corp in Boston, Massachusetts, which last year launched the Roomba, a robotic vacuum cleaner.
Robotics experts used to argue that if we were ever to have a single multi-purpose droid 鈥� complete, of course, with bow tie and a butler鈥檚 voice 鈥� it would need to have two legs to operate in the human-built environment. Two legs, it was thought, would get your robot to places that wheels could not, such as up and down stairs and high shelves.
But that idea is history, roboticists now say. Legs are no longer necessary. Tracked robots can handle stairs at more than useful speeds, and even if it were possible to program a robot to do all your chores, says Angle, you don鈥檛 need legs to clean a house or keep an elderly person company.
鈥淟egs decrease function, increase price and decrease reliability,鈥� he says. For example, it is much easier for a wheeled or tracked robot to carry a load than it is for a legged one. A biped needs continually to adjust the forces required to stay balanced, since its centre of gravity shifts with the tiniest additional load (see 鈥淎 question of balance鈥�). This is an enormous power drain. In addition, the motors are in constant use, even during moments of stability, resulting greater wear on their parts.
Power is the big issue, says Krupp, because battery technology isn鈥檛 up to the task of operating a robot for 24 hours without a recharge. Legs are simply too inefficient.
Even if Qrio could run about for longer than 30 minutes before conking out, there are many control aspects of walking that have yet to be solved. As impressive as they are to watch, the 60-centimetre high Qrio and Honda鈥檚 walker, Asimo, which is twice as tall, can only move on relatively even surfaces and have only a limited ability to adapt to changes in them. A kid鈥檚 toy on the floor could trip them up.
Angle puts the obsession with bipeds down to simple 鈥渟ex appeal鈥�. Walking is both an attractive and challenging way for Sony and Honda to parade their own techno-supremacy, hoping to give the impression that extreme high-tech will trickle down to their bread-and-butter businesses.
Honda says it recognises that the technology isn鈥檛 there yet. But ultimately it believes it can solve the power problem, and is aiming to create a useful multi-purpose android within the next 15 years. Meanwhile, the rest of the robotics industry will get on with designing more practical machines.
A question of balance
Despite the criticism from the robotics community, advances are being made in bipedal locomotion. Most two-legged robots use balancing software that follows a routine called Zero Moment Point (ZMP) calculation, which works out the forces and positions needed to keep a robot stable at all times.
But for ZMP to work the robot must be in contact with the ground at all times. Qrio鈥檚 advance is that when running, both its feet can lose contact with the ground without compromising stability, says Ken鈥檌chiro Nagasaka of Sony鈥檚 robotics division. And whereas walking involves constant control to maintain balance, running is in principle simpler to maintain because the robot鈥檚 greater forward momentum helps to keep it stable.
In the US, Hugh Herr, director of the Leg Lab at MIT in Boston, says his colleagues have just worked out a way to let a running robot cope with uneven surfaces. They have found that ensuring the leg is moving backwards as it touches the ground vastly improves stability.
