This site has been archived on
You are here: CORDIS / IST web / Emerging Technologies and Infrastructures / FET / Beyond Robotics (RO)

Future and Emerging Technologies

Find the most recent information on EU Funding activities in the field of Information and Communication Technologies (ICT) by visiting our ICT in FP7 website , which covers ICT in the 7 th Framework Programme (FP7) 2007 - 2013.

Find the most recent information on Future and Emerging Technologies activities in
the 7th Framework Programme 2007-2013 on our FET Proactive   - FET Open Web sites.

 

Beyond Robotics (RO) Proactive Initiative

Presentation of the Initiative (draft working document)

The purpose of this document is to provide some guidelines to potential proposers interested in submitting project proposals in the IST-FET proactive research initiative "Beyond Robotics", open in the IST-FP6 1st Call for Proposals (17 December 2002 to 24 April 2003). The document is structured as follows:

Section 1 recalls the objective of the initiative;

Section 2 discusses the main research challenges ahead and the specific research topics that need to be addressed;

Sections 3 - 8 present the scientific communities the initiative is addressing, highlight the relation of the initiative with previously launched IST-FET initiatives and provide some useful hints with regard to the types of project proposals to submit and their tentative budget.

  1. Objective
  2. Research Challenges
    1. Robust perception
    2. Learning
    3. Task and environment adaptation
    4. Interaction
    5. System issues
  3. Scientific communities addressed
  4. Link to the FET initiatives on " Neuroinformatics " and " Life-like Perception Systems "
  5. Composition of the initiative, financing instruments and indicative budgets
    1. Integrated Projects (IP)
    2. Networks of Excellence (NoE)
    3. Advisory Board
  6. Link to and co-ordination with national activities
  7. International collaboration
  8. Additional important issues to consider

1. Objective

Incorporation of information technology into embodied 1 artefacts ("robots") poses a wide range of interdisciplinary research challenges and has the potential to lead to a large variety of new applications.

The research initiative addresses one or more of the following long term objectives:

Cognitive Companions
The development of cognitive robots whose "purpose in life" would be to serve humans as assistants or "companions". Such robots would be able to learn new skills and tasks in an active and non pre-determined way and to grow in constant interaction and co-operation with humans.
A robotics "companion" can be considered as an adaptive servant that co-exists and continuously interacts with the user. For this purpose, it must evolve with the user so as to acquire the necessary skills, representations and competencies. As an example, it may be used as an "assistant" that is living and growing together with its master, is adapting itself to and is complementing the mental and physical faculties of its master and is eventually able to compensate for any degradation of its master’s performance. For the companionship, the robot will naturally be exposed to a wide variety of situations and events. Hence, the system must be able to acquire skills and perform new tasks. In addition, the constant exposure to new environments and settings imposes significant constraints on perception and reasoning capabilities to ensure robust real-time performance over years of use.
Human Augmentation
Hybrid bionic systems that would augment human capabilities such as perception of the environment, motion, interaction with other humans, etc. This would involve smooth integration of sophisticated robotic and information systems with human perception-action systems using bi-directional interfaces with the human nervous system.
For this objective, systems are to be developed in which there is a tight coupling between the human and the artefacts, e.g., intelligent prosthetics (artificial sensory organs, arms, limbs, etc). The system must be flexible enough in terms of physical interaction and skill/task adaptation to function as a regular member of the human body. High-fidelity, two-way brain interfaces or other more advanced interfaces could provide full support to accommodate human functionalities like manipulation or walking, but also for augmenting humans with new perception-action capabilities to go beyond our current physical limitations.
Robot Ecologies
The development of autonomous microrobot groups consisting of many heterogeneous members exhibiting collective behaviour and intelligence. The robots would be able to self-organise, adapt, co-operate and evolve in order to attain a global objective.
Microrobot groups can be viewed as parallel systems in terms of perception, reasoning and action generation that open up interesting issues for their co-ordination, adaptation and evolution in order to jointly accomplish a wide variety of different tasks. Given a large number of "rational physical agents", the issues of task distribution, co-ordination and sharing of knowledge are yet unsolved, in particular for heterogeneous systems that have to operate and co-operate in real-world, open-ended environments. This part of the initiative is closely connected to the simultaneous FET intiative on Complex Systems research .

Proposals should have ambitious objectives at the level of a complete system and aim at breakthroughs that go well beyond the state of the art. Research should be novel, long-term, visionary and seek new approaches. It should address and integrate topics such as novel sensors, multisensory perception, cognition, learning, task and environment adaptation, embodiment, interaction with humans, scalability, integration and rigorous evaluation. The work would partly build on the ongoing FET neuroinformatics (NI) and life-like perception systems (LPS) initiatives with emphasis on integration and systems research issues. The effort may start from or adopt existing state of the art solutions in robotics or robotic sub-systems, where appropriate. However, the research should not simply evolve along well established directions in areas such as Artificial Life, Artificial Neural Nets, Good Old Fashioned AI, flexible manufacturing and mechatronic systems, humanoid robotics, basic physical design or simulation only studies. These topics are relevant but on their own they are not sufficient to address the objectives of the initiative.

2. Research Challenges

The realisation of the above objectives implies highly interdisciplinary research work, and in particular, a close interaction between neural and behavioural sciences and information technology systems research.

The research effort should in fact focus on as many as possible of the following central research topics that are necessary to realising the objectives of the initiative:

2.1 Robust perception

Today, perception systems are vaunting successful applications in several sectors. Despite recent advances, there is still a huge need to come up with fast, robust and versatile perception systems that allow embodied artefacts to operate in unconstrained real-world environments; to endow systems with perception engines that, through multiple sensory modes, acquire, maintain and deliver selective knowledge rather than mere information and that are capable of recognising, making abstraction and categorising a large number of objects, contexts and events. In brief, there is a need to develop perception systems having true cognitive abilities that can be used for understanding situations and tasks and for interacting with humans and the environment.

2.2 Learning

Open-ended conceptual learning and skill, task and model learning

Despite the very active research efforts in the field, so far only limited results have been reported on conceptual learning and in particular on learning in the context of embodied systems. For such systems, learning has to be investigated at all levels, from learning of basic skills (fundamental perception/action primitives) to the development of a conceptual model of the environment (e.g., learning of maps, models and other entities) and to task level models, including basic parameter control, refinement and updating.

Learning capabilities of an embodied artefact need to be demonstrated when it operates in non predetermined, real-world environments. It needs to be shown also that the set of the artefact competencies is ever growable and evolvable, allowing for lifelong learning and use. A rich bouquet of learning techniques must be integrated to facilitate truly general-purpose learning.

2.3 Task and environment adaptation

Embodied and situated 2 artefacts are operating under diverse situations and under continuously changing environments and tasks. They must thus be capable of self-organising and self-evolving, so as to best adapt and accommodate changes. This calls for a high degree of flexibility and versatility both in terms of perception, reasoning and action but also of the artefact’s morphology.

2.4 Interaction

The degree of usability of an artefact and ultimately its future market success are critically depending on the flexibility and versatility of its interfaces. These need to be as natural and intuitive as the human-to-human interactions. As an example, direct interfaces to the nervous system or more traditional multi-modal interfaces could be considered.

Direct interfaces
For direct control of a device, brain interfaces need to be developed having short time delays (of the same order as in human interaction, i.e. about 0.5 s) and a bandwidth close to or better than 10 Hz. On the other hand, multichannel (n = tens to hundreds) invasive bi-directional interfaces are needed to mediate the finesse of human motor control and sensory inputs.
Multi-modal interfaces
When interacting with people in general, it is necessary that an artefact exhibits a self-evident and simple dialogue behaviour and has facilities to "explain" the interaction. These call for significant advances in multi-sensory perception, perception-action coupling, multi-modal interaction, cue integration, methods for extended dialogue behaviours and integration of "physical behaviour" with the more traditional interaction modalities, but also for an increased understanding of human emotional and cognitive processes.
2.5 System issues
2.5.1 Scalability (n>>1)
Scalability is a critical characteristic for the design of artefacts evolving in truly complex and dynamic environments. Existing robots are show only limited capacity in this respect 3 . They need to have scalability of several orders of magnitude beyond the one available today. For this, the following issues have to be addressed:

First, to develop methods for memory and knowledge organisation (categorisation) for handling information and knowledge accumulated over time and for performing their intelligent "forgetting" and abstraction (closely related to learning).

Second, to co-ordinate the diverse set of activities in a system and its interaction with the environment. In fact, to enable a more principled treatment of scalability implies making recourse to a theory of systems. Recent control theories can be adopted such as the control of hybrid dynamic systems integrating continuous and discrete events into a unified framework. These need to be extended to generalise to systems with hundreds of states. They also need to explicitly integrate issues of perception and knowledge and of their generalisation so that categorisation of objects and interpretation of events and situations can be dealt with, beyond their mere recognition.

In general, a new theoretical framework is required in which generalisation across tasks is inherently embedded and where computation, memory, perception, reasoning and action generated are integrated into a coherent and scalable fashion.

2.5.2 Open Architectures
Embodied and situated artefacts are inherently complex entities designed to evolve and adopt more and more varied forms of behaviour. There is then a strong need for developing system-level methods for design, analysis and deployment as well as schemata for overall systems optimisation and robust fault handling. The challenges include here building open architectures and frameworks for achieving adaptability, composability, distribution and concurrency of actions and tasks, evolvability, modularity of heterogeneous components and subsystems, coping with failures and for ever richer computing and sensing environments.
2.5.3 Integration
Integration in robot systems typically requires multidisciplinary knowledge of system issues from several different technical fields, e.g., signal processing, computer science, systems engineering, applied mathematics and control engineering. At least one of these fields is often compromised in the set-up of a system and thus system construction and integration is often viewed as a "black art". There is a strong need for adequate techniques to address integration in the context of theoretically comprehensive methods.

Furthermore, in truly multi-purpose systems operating in the presence of continuous task and environmental changes, it is impossible to identify in advance and embed in their design parameters copying with all possible situations that might be encountered. Deliberation and integration are key issues to facilitate design, implementation and deployment of such systems. There is a need for large interdisciplinary teams and for devising a new operational theory.

2.5.4 Test and evaluation
Most robotic systems of today have been evaluated in given application settings. Stringent scientific methods (similar to psychophysical methods in the behavioral sciences) have rarely been used for their formal testing and evaluation. Through the use of a solid theoretical basis, it is possible to formulate stringent protocols for a more thorough testing and evaluation of systems going beyond their assessment for demonstration purposes only and/or their qualitative assessment only.

The evaluation of systems should also include formulation of "golden" standards for rigorous testing and evaluation of performance in terms of perception, action generation, planning, learning and system operation. Through such standards, it becomes possible to perform comparative experiments that can illustrate clearly the strength and weaknesses of a system.

3. Scientific communities addressed

The "Beyond Robotics" initiative addresses a broad range of scientific communities and in particular:Communities working mainly in research fields such as in advanced robotics and control, learning, signal processing, perception and cognition, embedded systems and complex systems engineering and human-computer interfaces.

Work may also be required from communities in cognitive and behavioural sciences, cognitive neuroscience, experimental neuroscience, computational neuroscience, neuro-informatics as well as in biological cybernetics, neuromorphic-, biomorphic- and biologically-inspired engineering and bionics. For these disciplines, the initiative is also viewed as providing to them advanced experimental platforms for testing their latest scientific hypotheses and findings and as a source for further inspiration.

Other important scientific contributions are likely to come from the areas of artificial intelligence, mathematics and statistical data analysis, evolutionary sciences, cellular engineering, bio-mechatronics and micro- and nano-electronics.

A primary aim of the initiative is to bring together as many as possible from the above scientific communities to work effectively and closely together in appropriate configurations in order to best realise the objective and challenges of the initiative (as described in Sections 2 and 3 above).

The initiative calls in fact for a "collection" of all those multi-disciplinary efforts that are united by an overall research objective rather than their simple juxtaposition within an expanded project that is set of loosely connected objectives. It thus aims at encouraging effective communication between engineers, computer scientists, neuro-scientists and biologists and to move, whenever possible, towards entirely new concepts via close interactions between the respective scientific fields. Especially groups which do not work yet in the required interdisciplinary context are explicitly called upon to establish co-operation with the necessary counterparts in order to activate possible hidden potentials of novel interdisciplinary work, whenever these are dictated by the necessities of individual project objectives.

4. Link to the FET initiatives on "Neuroinformatics" and "Life-like Perception Systems"

The research topics of "Beyond Robotics" are linked to those addressed in the following two FET neuro-IT proactive initiatives: The " Neuroinformatics for Living Artefacts " initiative 2000, which views entirely grounded and situated cognitive systems and the " Life-Like Perception Systems " initiative 2001 that centres on perception-action integration. With respect to these initiatives, Beyond Robotics will have an augmented scope for learning, adaptation, scalability, integration and systems research. The development of entire autonomous robot artefacts will be necessary.

5. Composition of the initiative, financing instruments and indicative budgets

It is expected that the initiative will consist of:

  • two to three Integrated Projects (IP)
  • one Network of Excellence (NoE)
  • an advisory board.

The initiative will only use the two new financing instruments, i.e. IP’s and NoE. Depending on the quality and nature of proposals and the available total budget for the initiative indicative EC financial support per project can be estimated to be in the order of five to eight million Euro per IP and in the order of three to five million Euro per NoE.

5.1 Integrated Projects (IP)

This section is a complement to the "practical guide on the provisions for implementing IPs in the Sixth Framework Programme". It is strongly recommended to all interested IP proposers to carefully read this guide first. The guide may be found here .

In Beyond Robotics, each IP should focus on realising one or more of the three long-term objectives, i.e., Cognitive Companions, Human Artefacts and Robot Ecologies. For this, it will have to achieve a true integration of research right from the very beginning and bring together a critical mass in terms of intellectual capacity and experience from several different disciplines.

5.2 Networks of Excellence (NoE)

This section is a complement to the "practical guide on the provisions for implementing Networks of Excellence in the Sixth Framework Programme". It is strongly recommended to all interested NoE proposers to carefully read this guide first. The guide may be found here . In the context of a FET proactive initiative, the NoE would have a specific role: it would bring together the broader community active in the research domain of the initiative in order to provide a framework of co-ordination for discussing, exchanging and disseminating best experiences in this area; for co-ordinating research and training activities at the European level; and, for enabling the progressive and lasting integration of these activities. This may include the establishment of "distributed" centres of excellence, shared fabrication or experimental facilities, testbeds etc. It is further anticipated that such a network could include resources for launching open calls for proposals for "exploratory projects" that are similar in spirit and volume to the FET "Assessment Projects" of the FP5 .

The NoE will help elaborate and maintain a research roadmap in the area, in co-operation with the integrated projects, and it will also ensure a broad dissemination of research results emanating from the research in the initiative, stimulate industrial and commercial interest, and enhance the public visibility of the research. In order to achieve long-lasting integration between the research activities in Europe the NoE would seek co-operation with similar or related initiatives in the member states. It will also seek to establish international collaboration with similar initiatives at other non EU countries.

As an example, some of the possible activities are already addressed in the running FET-funded thematic network EURON .

5.3 Advisory Board

The advisory board will set up by the Commission as a part of the proactive initiative. It will consist of prominent scientists acting as the reviewers of the projects and the initiative as a whole, stakeholders from industry and other research disciplines, as well as selected representatives from national research programmes, projects or funding agencies (possibly also non-European) and the Commission. It will monitor the progress of the initiative and advise the projects and the Commission.

6. Link to and co-ordination with national activities

In order to achieve long-lasting integration between the research activities in Europe, the initiative would seek co-operation with similar or related initiatives in the Member States and the countries associated to FP6.

Each IP is then seen as being an integral part of this broader effort: each IP should then clearly demonstrate how their work is related to and co-ordinated with work carried out in national initiatives and show that they are building on top of or complementing work in national initiatives.

7. International collaboration

It is probable that an IP will consider to collaborate with and integrate in its team non-EU research entities. Such collaboration is possible when this non EU entity is bringing into the project critical knowledge and expertise not available within the EU and the countries associated to FP6, and which is considered essential for realising the project objective and goals. In most cases such an entity participates in the project at its own cost.

However, when well justified (e.g. the project would avoid spending resources in catching up the state-of-the-art in some essential sub-area of research) FP6 foresees international co-operation with the financial support to entities from non-associated third countries, provided that they belong to the categories defined in the Specific Programme 4 )

8. Additional important issues to consider

Test and Evaluation: each IP to be submitted in this initiative is required to address in its workplan testing and evaluation methodologies and protocols, as outlined in section 2.5.4 above. When doing so, it should also include in its workplan provisions for joint collaborative work with the other IPs and NoE that will be eventually launched within this initiative. The aim is to develop commonly agreed sets of system performance testing and evaluation protocols that span across individual projects and become the reference for the whole scientific field.

Explicit measures/benchmarks for system integration: An effort should explicitly specify measures or benchmarks for the overall project and the integration into systems. These measures will form a basis for evaluation of ambition, feasibility, and actual progress. It is thus crucial that such measures are explicitly defined as part of the project objectives.


1 "Embodiment" is the physical instantiation of an artefact. It is an aspect leading to specific considerations with regard to the limits and natural variations of the artefact’s sensors and actuators.
2 "Situatedness" is an aspect of an artefact signifying that it has to immediately respond to a current situation by optimally exploiting its (limited) resources.
3 For example, computer vision systems of today can at best recognise up to 20 different objects (by identity only rather than category) and, may be, up to 1000 different logos. Most mobile robot systems of today can carry out 10-20 different missions.
4 A budget of about 90 million Euro has been earmarked for participants from the following countries: Russia and Newly Independent states, Mediterranean Countries including the western Balkans and developing countries. Participants from other third-countries may also get funding in duly justified cases. For more details on the participation and financing possibilities by country of establishment of a participant, see Annex I of the practical guide on IP‘s .