The magic touch
Tom Shelley reports on why ‘feeling’ robots could outperform those solely dependent on vision systems
Future generations of robots could well ‘feel’ their way around rather than depending on vision systems and programming.
According to Professor Daniel Wolpert of the University of Cambridge Department of Engineering: “Tactile sensing is likely to be far more important than vision for robotics. I really believe that. Blind people can do a lot.”
Addressing a recent ‘Horizon’ seminar on ‘The Thinking Machine’ Wolpert pointed out: “Deep Blue can beat Gary Kasparov at chess; a five-year-old can place chess pieces on the board more efficiently and accurately than the best possible robot.”
He has therefore been engaged in studies of how humans learn to do such things and then “reverse engineer the human motor system”. This is more difficult than it was originally thought to be, taking into account the complex make-up of both the human body and mind: there are around 100 billion neuron cells in the human brain, each connected to some 7,000 other neurons, making 700 trillion connections.
Yet advances in machine vision recognition have been substantial, as Professor Roberto Cipolla, also from the Department of Engineering, demonstrated. Defence researchers, for example, have for decades been trying to develop systems that would differentiate trucks from tanks. This can now be done, mainly because of the gun barrels, but distinguishing a friendly tank from an enemy one cannot yet be done with any reliability. Meanwhile, in an industrial environment, there have been endless efforts to detect and identify mixed arrays of objects in various orientations, with varying degrees of success. However, says Cipolla: “It’s beginning to work.”
Vision systems are certainly good enough to distinguish known objects from one another and to locate objects at a distance, but are not presently of much use on their own when very precise control is required.
Wolpert explains that it is almost impossible to stand up, or even strike a match, without tactile feedback. This has been established from experiments and the experience of humans with disabilities. However, blind jazz musicians can become virtuoso players, provided they can feel the keys on their instruments.
There have been a large number of European Union funded projects on generating tactile feedback, mostly for remote sensing through haptic interfaces – but which could also be used to interface directly to machines. In a project called NanoBioTact: Nano-engineering Biomimetic Tactile sensors – announced at the end of 2006 and funded to the tune of €3 million – it is intended to develop an articulated artificial finger that could be connected directly to the central nervous system to give the user a sense of touch. The sensor is to be based on an array of NEMS force transducers intended to mimic the spatial resolution, sensitivity and dynamics of human tactile neural sensors. The project is being coordinated by Professor Mike Adams of Birmingham University and is due to be completed by the end of November 2009.
Another new project is called: Biotact – “Biomimetic Technology for vibrissal Active Touch” – inspired by the whiskers on rats and shrews. These creatures sweep their whiskers back and forth at high speeds, in order to establish the shape and surface of objects and capture prey. Coordinator Professor Tony Prescott at Sheffield University says: “Today’s life-like machines, such as robots, don’t make effective use of touch. By learning from nature and developing technologies that do use this physical sense, our researcher will be able to enhance the capabilities of the machines of the future.”
Other partners are: the Berlin Center for Computational Neuroscience; the Scuola Internazionale Superiore Studi Avanzati in Trieste; the Bristol Robotics Laboratory; the Ecole Polytechnic Federale de Lausanne in Switzerland; Ben-Gurion University and Weizmann Institute in Israel; and Brain Vision Systems and Northwestern University in Chicago. The EU is providing €5.4 million, out of a total project cost of €7.8 million.
* Touch will perhaps be even more important than vision in future robotic systems
* Tactile sensors normally detect pressure distributions, but there is also an idea being developed that relies on whiskers, inspired by rats and shrews
* Much tactile technology is already available, due to the amount of effort going into the development of haptic interfaces
This material is protected by MA Business copyright
see Terms and Conditions.
One-off usage is permitted but bulk copying is not.
For multiple copies contact the