ANNUAL MEETING CANBERRA 2 5 May 2000
Dr Alexander Zelinsky worked for BHP Information Technology as a computer systems engineer for six years before joining the University of Wollongong Department of Computer Science as a lecturer in 1984. Since joining Wollongong University he has been an active researcher in the robotics field. He obtained his PhD in robotics in 1991, then worked for leading robotics research laboratories in Japan as a research scientist from 1992 to 1995. In 1995 he returned to the University of Wollongong Department of Computer Science as a senior lecturer. In 1996 Professor Zelinsky joined the Australian National University Research School of Information Science and Engineering as Head of the Robotic Systems Laboratory, where he is continuing his research in robotics. In 2000 he was promoted to Professor and Head of Systems Engineering. He is currently president of the Australian Robotics and Automation Association.
The 21st century: The age of intelligent machines
by Alex Zelinsky
In the last 40 years we have seen the development and growth of the robotics technology field, which started with simple pick-and-place devices with little or no sensing capabilities and has now progressed to the development of machines that can dynamically balance and walk like humans, and that have sophisticated sensing capabilities such as touch, sound and vision. Robotics technology is now reaching the point where it is ripe for large-scale deployment in everyday, real-world applications. Professor Zelinsky will describe the likely developments in robotics research and development in the next 10 years, and outline the drivers (industries and disciplines) for robotics technology development. He will also speculate on the possible commercial and scientific opportunities that robotics could bring to Australia.
We’ve all seen Hal in 2001: A space odyssey. Can we do this?
What is a robot? Mike Brady from Oxford University has said, ‘Robotics is the intelligent connection to action.’
The dream is the droids from Star Wars. The reality is a mobile arm shifting boxes.
The Japanese have made a robotic pianist – a hand with fingers that strike the keys. But it is a dedicated machine that cannot be reprogrammed.
Honda has built a humanoid robot that walks down stairs and can resist pushing. The Massachusetts Institute of Technology has built machines inspired by biology that run, hop, walk. Others have built a hand that can dexterously manipulate a tin. Japanese robotic dogs sold thousands in a few days.
A female roboticist taught three robots to dance, communicating, moving in harmony, working with people.
Future research can go in a number of directions. To build humanoids like the Honda robot we need to learn more about human biomechanics. To mimic human intelligence we need to learn how the human mind works – the visual system, learning, memory and consciousness. A robot that learns would be really useful.
But why do robots have to be like humans? A machine that can go beyond human capabilities – that is larger, smaller or can go to extreme environments – could have many field and service applications – underground mining, land mine clearance, space exploration (astronauts don’t like robots!), surgery, under the sea, in the air, in agriculture.
Escaping from the limitations of humanoids is like solving the problem of human flight with aeroplanes rather than mechanical birds.
The research questions are primarily engineering. Individual robots are still blind, so we need to learn more about perception, especially vision. There is a race to improve this technology.
In the area of mobility, a new mechanism can make the robot more useful and save energy. One of the big problems, as for electric cars, is the power supply.
Because it is impossible to program a machine for all possibilities, the system has to be able to learn.
Robot work cells are dangerous. So we need to find ways for humans and robots to coexist.
A number of fields are driving this technology. The military, which pushed computers and the global positioning system, has many potential applications. Primary and secondary industries could use military spin-offs for dirty, dreary and dangerous jobs. The entertainment industry uses a lot of robots, for example dogs and other animals in movies. And as health care costs rise, robots could be used to help sick people stay at home longer.
All disciplines can contribute to robotics. Systems theory can devise methods of safe operation. Biological technology, for example Srini’s work with bees, will help. In Japan, scientists drawing on medical research have combined living tissue with a robot.
Australia still makes its living as a quarry or a farm. The field of robotics will be vital to maintain productivity and efficiency in these industries. But it also offers a niche export opportunity. Australia has world leaders in areas of research that could add value to existing machinery. Australia has to survey its oceans; these are so vast that the survey will need to be automated.
Will robots create unemployment? Robotics creates jobs if the industry is in Australia. At the moment the highest number of robots are in the nations with the lowest levels of unemployment. If we can make robots for the consumer, that will create a lot of employment.
Robotic dogs create a moral dilemma. Is there any danger of humans having wars with robots?
Alex Zelinsky. Will robots take over the world? I have always been suspicious of robotic pets. They may be okay if you look at them as educational devices. But toys can teach the wrong things. Robots are tools that can help you do your job. We are a long way from wilful conscious robots; they cannot yet surprise the designer.
In terms of hardware manufacture, Australia is the poor man of Asia. In 1998 the head of Intel spoke to the Federal Cabinet. Intel built a $2 billion plant in Israel. What is the future of billion dollar activities in Australia?
Roger Kermode. We’ve had the lead, with CSIRAC and the black box flight recorder, but we didn’t exploit them. Companies go where it’s easy to build and get to market. There is also value in ideas that lead to products. If there is a choice between manufacturing and intellectual property, I would rather have intellectual property. Shipping costs from Australia are prohibitive.
Alex Zelinsky. Australia was making computers long before Taiwan. Taiwan now has a computer industry 100 times bigger than ours. We can’t lament opportunities missed, for example mobile phones, we need to look ahead. The application of robots to a field is what we should be looking at. Learn from the past. We should manufacture high-tech big-margin products, not appliances.
Terry Percival. Remember the value chain. Manufacturing is 10 to 30 per cent of the value of a product. We’re working to create intellectual property. I used to build things, now I do it in software. One program, which cost $1 million to develop, brought in $10 million last year for its intellectual property, with nothing manufactured.
How is the power problem in robotics being dealt with? Can renewable energy help?
Alex Zelinsky. The big problem is that robots consume a lot of energy. Solar cells sufficient for a humanoid would cover the Academy’s Dome. Most motors are electric: portable electricity depends on battery technology. There is a lot of work being done on that. Batteries can last a few hours but then they need to be charged. For a mobile robot there are no power points in the field.
Biological power supplies are very efficient: with one kilogram of food a human runs all day. Isn’t this what we should be looking for?
Alex Zelinsky. Yes.
The technology for virtual enterprises already exists, but it can only be marketed to large companies. How is the Internet going to make this technology more accessible? Or will the technology provide a technological advantage to large companies?
Terry Percival. It will be an advantage to large corporations. But smaller companies will link into their systems. Large companies get unwieldy; than small companies can take advantage of the technology.
As geography becomes less important, Australia has an enormous opportunity to reach new markets we couldn’t look at. What areas of business should we be getting into?
Terry Percival. Software is one. India has huge software companies up and running. But Australians don’t buy Australian software. Networking technology offers a huge opportunity. In Canada Nortel took on 10,000 new researchers last year.
Roger Kermode. Motorola has 300 software engineers in Adelaide, and researchers producing intellectual property in other cities. It’s all pure export business.
Anything that removes the middle man is where the big opportunities are. People don’t care where software comes from, they just download it from the net. Getting the inefficiencies out of business is a big thing.
Where does Moore’s law run out of steam and hit an asymptote? When will the physical limitations prevent further advance?
Roger Kermode. Five years ago people said five years. Now people say five years. We don’t really know. Quantum computing is now being explored.
Bob Frater. Moore’s law reviewed different technologies. In the next 15 years, chip makers will find ways to hold to the curve. They have to in order to survive.