The Robots are coming!

Jon Excell visited android designers Shadow to find out just when we’ll be able to give up the house work and do something more interesting with our spare time

The notion of building a humanoid robot which can cook, clean, iron, sew and do the gardening for you has, in a sense, become quaint; one thinks of Robbie the Robot or mankind’s arch enemies, the Cybermen. We still assume that one day this will be possible, and every now and again the collective imagination is fired up by Hollywood, but, for the main part, robotic development is concerned not with one machine that can do everything, but with the pick and place machines we see on the factory floor.

Thus, large scale development of the domestic android is surprisingly uncommon, and the Shadow Robot Company, a small network of dedicated enthusiasts working from a workshop in North London, is modestly bewildered to find itself voted fourth in a list of the world’s greatest android projects, hot on the heels of giants like Honda.

Established over 10 years ago in MD Richard Greenhill’s attic, Shadow’s aim is to design a robot which imitates the human body as closely as possible. DE visited Shadow at its workshop, and asked, how long will it be before we have a domestic android?

Surprisingly, Artificial Intelligence (AI) is such a heavily researched subject that it’s one of the smaller hurdles in the design of the android. AI breaks down into two areas: being able to function in the world and having a more abstract creative ability. Quoting Moore’s law, (which states that the power of computers, governed largely by transistor density, doubles every 18 months) some say that the computer will soon rival the human brain. But, as Rich Walker – technical director of Shadow – points out, not only would a machine with this kind of processing power be incredibly difficult to program, it would also be unnecessary.

We need to ask ourselves what we want these assistants to be capable of, and Walker’s answer is that they must be able to perform repetitive tasks and build a representation of the world, not possess the ability to play the cello and write sonnets.

Accepting this aim, the technology is already well advanced, and Shadow, with its limited resources, sensibly leaves this area of development to the IT companies who have the time and money to get to grips with it.

In order to build a three dimensional view of the world and carry out commands, the robot must also be equipped with a vision system and speech recognition software. Again, these are areas of technology which are well developed. Speech recognition software is becoming incredibly advanced. Walker recounts an anecdote about someone shouting in a crowded room, at a computer running voice recognition software, `Format C…Colon….Yes…Yes’. The machine promptly formatted its own C drive. “You will interface with a robot verbally,” adds Greenhill with great certainty.

Vision systems again are largely the domain of big organisations. For example, scientists at Johns Hopkins University in Baltimore have developed an integrated low-power chip which, as well as imitating the operation of the high-definition region of the mammalian retina, also has a peripheral vision area which tracks the location of objects. Greenhill feels that Shadow should be doing more in this area, but research in vision is a modular process, where someone might spend three years studying the way a gaze moves around a room. Greenhill’s frustration is tempered by this knowledge.

The robot’s spatial awareness will not, however, be determined purely by vision. When it encounters objects outside its field of vision, the `central nervous system’ will take over. Sensors on the robot’s body will report back to the computer that something unexpected has happened and the robot will stop moving. It will attempt to find out what has happened and, if it doesn’t understand, the most important reaction, says Walker, will be for it to realise this, and ask for help.


What stands out about these technologies is that they have not been designed specifically with the humanoid robot in mind. VR software is used in all manner of applications, the `Eye’ chip will be used in toys or monitoring devices and even AI is rarely associated with androids. Shadow keeps a close eye on what is going on and in the meantime concentrates on its own area of expertise; building the hardware, the body, which will embrace with open arms the technology which gives it `life’.

So why should a domestic android mimic the form of a human? The robot must exist in our world and work within our ergonomics, while four legs may offer great advantages in terms of stability, they would be impractical in the home. Also, the human body is there as a copyable model which Shadow intends to exploit.

The project’s chief breakthrough has been the development of its air muscle (see DE May 1999, page 68), a technology which Greenhill feels is `essential’ for robotics. Prior to this, robotic movement relied on either motors, hydraulics or pneumatic cylinders, all of which have disadvantages for this application. “Hydraulics are expensive, heavy and unsafe for domestic use, pneumatic cylinders tend to stick and motors usually go too fast to simulate natural movement in nature,” says Greenhill. “Very little goes round and round; nature is built with levers.” And this is why air muscles fit Shadow’s plan perfectly. Not only are they light, inexpensive and safe, but they mimic human muscles and run on air, a hopefully endless resource.

Using air muscles, Shadow has built an arm and shoulder – which demonstrate almost the full range of human movement – a hand, and most impressively, a bipedal robot (pictured in DE May 1999, page 68) with 14 muscles and 50 sensors on each leg. Lots of people have built bipeds, many of which are capable of taking more than the few stumbling steps completed by Shadow’s creation, but none, claims Walker, move like humans. “Walking as humans do it is a process of falling over and catching yourself; walking as standard robots do it is a process of standing still and moving the other foot – balancing on one leg at a time,” says Walker. Shadow’s biped stands like a human, and when it walks it falls and catches itself, like a human. The similarity to our own physiology is eerie.

The Shadow Robot Company is motivated by many factors: the desire to improve the quality of life of disabled people, the desire to change our lives, the desire to get there first, and a passion which one feels has been with its members since infancy.

The company claims that if it had $100 million, within 5-10 years it would have a robot which could perform most household chores. And this is believable. After all, Shadow’s biped was built for only £2000 (a fraction of the cost of Honda’s robot). The technology for the design of a humanoid robot exists; it is really a case of bringing it all together. Perhaps, hints Chris Phillips – Marketing Director – we would have domestic androids already were it not for the “money and will being taken out of robotics during the Thatcher years.”

And so, we must wait, and endure a little longer those boring household chores which stop us from playing the guitar, or reading books, or playing sport… Or building robots.

SHADOW: Tel: 0171 700 2487