Now for the human robot

At 73 Dr Joe Engleberger has a lot of living left to do. In the US, where he lives, the fastest growing age is 85. The visionary who is credited with putting robots into factories all over the world has turned his back on the industrial scene, and is developing a robot to enable people […]

At 73 Dr Joe Engleberger has a lot of living left to do. In the US, where he lives, the fastest growing age is 85.

The visionary who is credited with putting robots into factories all over the world has turned his back on the industrial scene, and is developing a robot to enable people to stay at home when they become infirm (The Engineer, 22 May).

He says he must look ahead to when, mind still intact, his body will need help to do the everyday things most of us take for granted.

Engleberger, a high-profile pioneer, is not the only engineer to anticipate a lucrative global market for robots in the service sector.

Industrial robots form an industry worth $8bn a year. Mobile service robots, in the home and public places, will become commonplace, Engleberger says. He expects a huge impact on the market.

So far his company, HelpMate Robotics, has put 120 mobile robots into some 97 hospitals around the world, helped by strategic partnerships including one with Otis Lifts in Europe and another with Yaskawa Electric in Japan.

Fourteen patents on the hospital robot highlight the wide range of technologies that are under development, including sensing, vision, speech, cognitive and intuitive as well as electro-mechanical and materials systems.

Engleberger’s latest and biggest venture, the safe and caring domestic robot, represents robotics’ most significant challenge to date.

Interest in robots for the service sector regained momentum 18 months ago when Honda launched a battery-operated robot capable of making decisions about its terrain. The autonomous biped climbs stairs and slopes, resists falling over when pushed and changes direction depending on conditions without being programmed to do so.

It was developed at Tokyo’s Waseda university at a cost of $15m $500m, is made of metal and weighs the same as a small car. It will never be generally used because of its weight, cost and hazard potential. But it is an effective technology demonstrator and sets new goals.

Many researchers argue that to perform human tasks well robots need to be lifelike. There are six to 10 humanoid robot projects in the world and 20 25 centres researching them.

In the UK, the leading exponent of robots with a human form is the Shadow Robot Project in London. Rich Walker, a mathematician from Cambridge University and Shadow’s software specialist, explains that human-style limbs are necessary for domestic robots to climb stairs and do other things to avoid the alternative, which is to replace houses with bungalows.

‘We took a conscious decision to keep the human musculature even though it is more difficult, because we figured out it must be like that for a reason. Unless we find a good reason not to do so, we keep to the human design,’ he says.

One exception is the robot knee cap. A force of between a third and half a tonne exerted by the robot’s pneumatically operated muscles is ‘horrendous’, says Walker. A pulley system which acts like tendons across the knee joint reduces that force to a few hundred kilograms.

Shadow’s anthropomorphic approach is controversial. Team members also shun measurement and dimensional drawings.

Greenhill, an inventor with no formal scientific background, has recently developed a robot hand that can move like a human’s with the same strength and degrees of freedom of movement, which he believes is a unique achievement.

He modelled the hand on his own, carving the skeleton freestyle from maple which is strong and has a similar density to bone, using traditional woodworking tools. He expects to demonstrate the hand in public later this year.

In Shadow’s domestic robot, the Biped Walker, joint movement is actuated by so-called air muscles designed, developed and patented by Greenhill.

These biologically inspired devices, which weigh the same and act with the same force as human muscles, are light, low-cost and safer than electrically operated actuators in many applications.

Referring to the Honda robot Dr Darwin Caldwell, head of robotics at Salford University, makes the point about weight.

‘If it stamps on your foot it will break it. If it falls on you, it will kill you,’ he says.

The Honda robot’s movement is controlled by electric motors, and it weighs 210kg.

In contrast, the humanoid robot at Salford, which uses air muscles developed by the university, weighs just 21kg. Researchers aim to produce a general purpose robot that learns by its own experience. ‘At present it walks like Frankenstein’s monster,’ says Caldwell.

Shadow’s Biped Walker has 28 air muscles which are controlled remotely by computer. Walking, a form of controlled falling, is said to be easier to achieve than standing still.

Compressed air at 2bar is supplied to the Biped Walker’s muscles through washing machine valves and the pressure is regulated using a standard Bordon pressure gauge. Low-cost strain gauge sensors in the muscle can detect movement down to 0.1mm in a spring which connects the muscle to the limb.

Shadow, like most leading-edge researchers, has little funding and relies on surplus store supplies and ingenuity for much of its development work. Income is generated by consultancy and sales of air muscles, which are hand-made.

Recently, the group has attempted to become more commercial by setting up a company and appointing a management consultant to the board. Project work includes aids for the disabled, nuclear plant decommissioning and minefield clearance.

Air muscles are different from most other actuators such as electric solenoids or mechanical air cylinders because they pull like human muscles, rather than push.

Usually they operate in pairs to control the movement of joints. One muscle is in tension while the other is in compression. To work, a muscle starts in tension as an elongated rubber tube.

As the tube fills with air, it expands outwards and contracts in length. When the air pressure is released, a braided outer sock compresses the tube, forcing the air out again.

Different sizes produce different forces. The smallest 12mm diameter muscle pulls 78N while the biggest, at 30mm diameter, will lift 3200N.

Simple and effective, Shadow’s air muscles are being used by US space agency Nasa in its telerobotics programme. Most recently, the Japanese have shown an interest in the device. Professor Hirochika Inoue of the University of Tokyo department of mechano-informatics visited Shadow in March and placed an order for air muscles worth £1,000. They cost between £5 and £55. Japan has begun a big humanoid robot project with government funding from MITI, the equivalent of the UK Department of Trade and Industry.

Robot intelligence is stretching the minds of all the researchers. Engleberger will look to Nasa to give his £3m domestic robot a brain.

Rich Walker at Shadow has moved away from fuzzy logic solutions which, he says, provide an unnecessarily high level of intelligence to lower level neural networks which are more akin to the ‘hardwired’ neurons in the human spinal column. Fuzzy logic is slow and hard to train ‘because you need a rule for each case’, Walker says. ‘Standing up by reasoning about it is basically a non-starter,’ he adds.

At the University of Reading, Dr William Harwin and his team in the department of cybernetics are using genetic algorithms and PC programming to give their biped robot intelligence. They are studying what they call pseudostatic walking gates to get the robot to control its own centre of gravity, which is one way of stopping it from falling over.

‘The end point is to develop machines that can learn. We would like robots to learn from their experiences as well as from instructors,’ says Harwin. That will simplify the design of future robots.

Robot researchers agree their work is underfunded, given the impact they believe a general purpose robot will have on the economy. In the mid-1980s the DTI invested about £150m of public funds on the Advanced Robotics Initiative. The focus on pure research is now considered to have been misguided because it did not include end users in the early R&D work. As a result there were no customers.

In 20 years’ time, general purpose robots may be commonplace. But researchers need to be patient, says Engleberger. He has learned from his experience of developing and marketing industrial robots for years without profit.

The breakthrough came in 1975. ‘It was a long haul getting there. Many people at the leading edge have the problem of convincing others. It takes a long time,’ he says.