This section has been transferred to the following http://ec.europa.eu/digital-agenda/en/science-and-technology/robotics . Please update your bookmarks.
ICT Challenge 2: Cognitive Systems and Robotics
Note: This is a short overview of targeted activities and is not legally binding. Fuller details on targeted activities are provided in the multi-annual ICT work programmes which can be downloaded from this site's Library section .
Why is it important?
Providing the next generations of ICT systems and products with more intelligence will open the door to a wide range of opportunities for ICT-based applications in a range of sectors.
ICT systems - including robot and robotic systems - need to be more robust, context-aware and easy-to-use. Endowing them with advanced learning, cognitive and reasoning capabilities will help facilitate this. Such capabilities should help them adapt to changing situations as well as to user needs and preferences. They should also help them to carry out tasks intelligently with people, and in the service of people, for example as assistants in the workplace, as home companions and servants, or as automatic agents for security.
Where do we stand?
Computing devices are supposed to make all sorts of machines and systems more intelligent and usable: on shop floors, offices and power plants, and in everyday environments. However, without extensive human intervention these machines and systems do not usually adapt to changing service requirements. Often, their performance degrades or they simply stop functioning - even in response to minor changes in their operating environment. Their ability to learn from past experience and to improve their services is minimal, if present at all. And they are not particularly user-friendly.
Robots are a case in point. Nowadays they are used mainly in manufacturing and engineering environments. However, new markets, such as home care and medical intervention, rescue and retrieval operations, and entertainment, will open up for robots and robotic devices if we can overcome limitations in versatility, robustness and their ability to interact with people.
With today's methods and technologies computing devices are largely confined to situations that can be unambiguously specified. The context- and user-awareness of machines, and systems based on them, are still wanting. For example, we still have difficulty representing a car so that an artificial vision system can always detect it as a car, irrespective of type, view angle or lighting conditions; it is equally difficult to have such a system recognise the typical behaviour of, say, a potential trouble-maker in a crowd, and to identify the person. Here, our lack of methodical knowledge severely limits not only the applicability of visual or multi-modal surveillance systems in public places but also hampers the proper use of repositories of still and moving imagery, as available for instance on the Internet.
The cognitive load on people dealing with systems of ever growing complexity is indeed enormous, in particular when safety or security are at stake. To lessen this load these systems must themselves become cognitive. They must be able to understand events, objects and people in their environment - and they must be able to understand their own situation and their own limitations. Current means of machine-mediated interaction and communication fall short of satisfying this need.
Where do we want to go?
In the real world, ICT systems should be able to respond intelligently to gaps in their knowledge and to situations that have not been specified in their design.
They should be able to exhibit robust and versatile behaviour in open-ended environments, give sensible responses in unforeseen situations, and greatly enhance human-machine interaction.
We therefore want robots to understand their environments and their users while operating either fully autonomously or in cooperation with people in complex, dynamic spatial environments.
We want artificial systems that can understand and control material and informational processes e.g. in industrial manufacturing or public services domains, for instance through real-time information gathering and interpretation in natural or artificial environments.
We want artificial systems to allow for rich interactions using all senses and for communication in natural language and using gestures. They should be able to adapt autonomously to environmental constraints and to user needs, intentions and emotions.
What impact do we want to achieve?
Meeting this challenge will provide enabling technologies that apply across domains such as robotics and automation, sensing and process control, complex real-world systems, image recognition, natural language understanding, and automated reasoning and decision support.
Apart from advancing new engineering methods and their scientific foundations we expect the work in this area to have significant industrial and societal impact.
Work will extend the industrial robotics market to:
- flexible small scale manufacturing, opening up professional and domestic services markets to robots,
- novel functionalities for embedded systems,
- assistive systems for interpersonal communications, such as support of dynamic translation, and effective medical diagnostics and therapeutics.
It will extend the capabilities of people to perform routine, dangerous or tiring tasks in previously inaccessible, uncharted or remote spaces; or help save crucial time in emergencies or hazardous situations.
A new approach is needed
Meeting this challenge will require rethinking the way we engineer systems. As we aim to get machines to exhibit performance capacities that resemble those of humans or animals, inspiration and insights will be borrowed from bio-sciences, social sciences and humanities. This may include the study of new computational paradigms derived from models of natural cognition. In some domains the exploration and validation of the use of new materials and hardware designs is strongly encouraged. Engineering progress will crucially depend on advancing our scientific understanding of what both natural and artificial systems can and cannot do, and how and why.
Structuring research and development in relevant areas, addressing requirements for cognitive capacities of artificial systems and ways to meet these requirements, is therefore a key part of this challenge.