european robotics research network

7/31/2019 European Robotics Research Network

1/137

FP6507728

EURON

European Robotics Research Network

Network of Excellence

Information Society Technologies Priority

DR 22.3

Special Interest GroupsYr 4

Due date of deliverable: 30 April 2008Actual submission date: 2 June 2008Start date of the project: 1 May 2004 Duration: 48 monthsOrganisation name of lead contractor for this deliverable: KUL

Revision: First versionDissemination level: PU


2/137

Two Special Interest Groups (SIGs) were funded during the Year 4: Cooperative Robotics and GoodExperimental Methodology. The latter one was a new SIG, stimulated to become active by a doublefunding opportunity from EURON, because its subject was to be a focus point of Call 3 of the EuropeanCommission, April 2008. Both SIGs lived up to the expectations, without much further interventionfrom the Board (it only had to remind the SIG Coordinators of their roadmapping and reportingresponsibilities). The Cooperative Robotics SIG presented its roadmap under the assumption thatthe EURON network would receive an extension covering the Summer of 2008; they will finish this

roadmap without the further financial support of EURON. Both SIGs succeeded in gathering criticalmass in their meetings; their detailed reports follow later in this document.

2


3/137

EURON SIG on Good Experimental Methodology Final Report

EURON

European Robotics Network NoE

Special Interest Group

GOOD EXPERIMENTAL METHODOLOGY

Final Report

Author: F.P. BonsignorioVersion: 0.9

Abstract

This document describes the work done so far within the SIG on Good Experimental Methodologyfunded by EURON Network of Excellence.

1/10


4/137


Table of Contents

Abstract.................................................................................................................................................1SIG Description....................................................................................................................................3Partners.................................................................................................................................................4Budget...................................................................................................................................................6

Organization ....................................................................................................................................6SIG Deliverables..................................................................................................................................8SIG Evaluation Criteria......................................................................................................................10

2/10


5/137


SIG Description

All science proceeds from experiment, which motivates the creation of new theory and establishesthe limits and validity of the existing theoretical basis. Individual branches of science conductexperiments differently, depending on the topic of investigation, but all have in common a body ofknowledge concerning experimental methodology that specifies how to design and conduct `good'experiments in that discipline.

If robotics aims to be serious science, serious attention must be paid to experimental methods. Wecan all point to published papers that report poor experimental work: few or no replications or tests;no, or dubious, comparison between algorithms; lack of suitable quantification of performance andits variability; conclusions which, while perhaps correct, are unjustified by the reported

experimental work. It is time to address this.

Our GEM (Good Experimental Methodology) Special Interest Group has the focus of increasing thequality of experimental methodology practiced in robotics. We believe this general aim can beachieved, for instance, by sharing good practice via educational workshops, summer schools, emaildiscussion and web presentation; by providing assistance to journal and conference reviewers andeditors concerning what constitutes experimental robotics and good practice in that sub-discipline;

by encouraging the principled replication and comparison of results; and by encouraging thedevelopment and use of appropriate systems benchmarks and standard evaluation procedures.This was going on during the EURONII funding and will continue as one of the activities ofEURON III

There is a clear community interest in these issues: in response to an earlier email circulated oneuron-dist some 80-90 positive responses were received.

The SIG has a small but active Board who maintained and is maintaining momentum and attempt toengage the large, interested, but busy community in improving the quality of our experimentalmethodology. We see benchmarking as an important part of good experimental method andencourage that activity within the SIG. The SIG is focused on experimental methodology with aspecial emphasis on replication and benchmarking. We do not underestimate the importance of

purely theoretical research and related papers, we want to define clear rules for experimentalpapers.

3/10


6/137


Partners

Coordinator: Fabio Bonsignorioof the Heron srl RTD group of EURON member 153, Heron srl

Partner 2: John Hallamof the Adaptronics Group group of EURON member 16, University of Southern Denmark (SDU)

Partner 3: Angel P. Del Pobilof the Robotics Intelligence Lab group of EURON member 3, Universitat Jaume I

Partner 4: Amiram Moshaiovof the Faculty of Engineering group of EURON member 21, Tel Aviv University (TAU)

Partner 5: Alessandro Saffiottiof the Centre for Applied Autonomous Systems (AAAS) group of EURON member 28, rebro

University

Partner 6: Cesary Zielinskiof the Institute of Control And Computation Engineering group of EURON member 22, Warsaw

University of Technology

Partner 7: Nikos Vlassisof the Intelligent Systems and Robotics Lab group of EURON member 182, Technical University

of Crete

Partner 8: Olivier Michelof the Cyberbotics ltd group of EURON member 128, Cyberbotics ltd

Partner 9: Danica Kragicof the Computer Vision and Active Perception Lab group of EURON member 1, Royal Institute of

Technology

Partner 10: Andrea Bonariniof the Dept of Electronics and Information group of EURON member 89, Politecnico di Milano

Partner 11: Javier Minguezof the Computer Science and Systems Engineering group of EURON member 71, Universidad de

Zaragoza

Partner 12: Domenico Sorrentinoof the IRA Lab group of EURON member 96, University of Milan Bicocca

Partner 13: Matteo Matteucciof the Dipartimento di Elettronica e Informazione group of EURON member 89, Politecnico di

Milano

4/10


7/137


Partner 14: Dirk Spennebergof the Mathematics and Computer Science Department group of EURON member 60, University

of Bremen

Partner 15: Paolo Darioof the ARTS Lab group of EURON member 101, SSSA

Partner 16: Alexander Brandleof the Intelligent Environments Group group of EURON member 173, Microsoft Research Limited

Partner 17: Marcus Vinczeof the Automation and Control Institute group of EURON member 10, Vienna Univesity of

Technology

Partner 18: Juan Domingo Tardos

of the Robotics, Perception and Real Time Lab group of EURON member 71, Universidad deZaragoza

Partner 19: Antonio Bicchiof the Centro Interdipartimentale di Ricerca "Enrico piaggio" group of EURON member 92,

University of Pisa

Partner 20: Maarja Kruusmaaof the Intelligent Materials and Systems Lab group of EURON member 154, Tartu University

Partner 21: Herman Bruyninckxof the Production Engineering, Machine Design and Automation group of EURON member 11,

Katholieke Universiteit Leuven

Partner 22: Gurvinder Singh Virkof the School of Mechanical Engineering group of EURON member 145, University of Leeds

5/10


8/137


Budget

The max available budget was 30k .Only about 15-20 k (precise numbers soon) has been spent.That happened for various reasons:

1) we expected an euron extension to the end of 2008, although we organized a comparativelyhigh number of meeting and workshop for a few months time frame (8, see below), thisdidn' t top the budget max estimate. Such an extension would have allowed an even majorimpact and involvement of the related communities

2) A number of people attending the workshops used other/their own funds and didn' t claimreimbursement this seems an indicator of interest

3) the expenses were managed carefully with a clear focus on value/price ratio.

The work volunteered by the proposers or other people who may contribute to the deliverables.The reimbursement was given upfront the receipts of the sustained costs (copies accepted when thereimbursement was not full).

Organization

Board

Fabio P. Bonsignorio: CoordinatorJohn Hallam: Co-chair Experimental MethodologyAngel P. Del Pobil: Co-chair Benchmarking

Proposers

Amiram MoshaiovAlessandro SaffiottiCesary Zielinski

Nikos VlassisOlivier MichelDanica KragicAndrea Bonarini

Javier MinguezDomenico SorrentinoMatteo MatteucciDirk SpennebergPaolo DarioAlexander BrandleMarcus VinczeJuan Domingo TardosAntonio BicchiMaarja KruusmaaHerman Bruyninckx

Gurvinder Singh Virk

6/10


9/137


80-90 people in Euron expressed their interest.The SIG is open to anyone who can and wish to contribute.

7/10


10/137


SIG Deliverables

D1) A GEM Guidelines document enumerating recommended quality criteria for Robotics Journal/Conference reviews that the community expects a proper high-quality experimental robotics paperhave to satisfy. This could be formulated as a reviewer checklist and circulated to Journal Editorsand Conference Programme Committees. We could also use the criteria when reviewing papersourselves. For instance, EURON might adopt of these rules in EUROS08.

D2) A web facility that includes information on planned meetings and SIG activities, educationalmaterial on good experimental practices, recommendations and outcomes from SIG/communitydiscussion, and a compendium of other efforts to promote good experimentation and quantitative

methods in robotics, both in the EU and internationally.

D3) We have organized/supported workshops on GEM and Benchmarking at major roboticsconferences We have organized open meetings preferably coinciding with the major roboticsconferences. Those meetings will continue after the EuronII deadline.

List of the meetings

Benicassim, Spain,9/1/2008, before, colocated with RISE08Genova, Italy, 31/1/2008,Zaragoza, Spain, 4/3/2008

Prague, 28/3/2008, part of Euron AGMValencia, Spain, 30/4/2008

Two GEM SIG meeting are scheduled beyond Euron funding period in 2008:

Zurich, CH, 27/6/2008, before colocated with RSS08

Nice, France, 25/9/2008, before colocated with IROS 2008

List of the workshops:

San Diego, USA, 2/11/2007, IROS 2007Prague, Czech Rep., 25-26/3/2008, before/colocated with Euros2008 and Euron AGMKarlsruhe, Germany, 1/4/2008, before colocated with CogSys08

Workshop are scheduled beyond Euron funding period in 2008:

Zurich, CH, 28/6/2008, RSS08

Nice, France, 26/9/2008, IROS 2008

A special session at Clawar 2008, Coimbra, Portugal 8-10/9/2008 will be held, another one has

been proposed after an invitation to ISR 2008, Coex, Korea, 15-17/10/2008

8/10


11/137


Other outcomes of the SIG have been so far:

1. links with related efforts (e.g. RoSta, PerMIS, Rawfeeds, Euron Benchamrking WP) ;2. community-endorsed examples of best practiceguidelines on selected subtopics3. Encourage and facilitate the publication of replications of robotics results, either in existing

robotics journals or by establishing the Journal of Replicated Robotics Results (a highquality open access web-based journal that specifically encourages the publication ofreplications of published experimental results) The first step is the submission to IEEE T-RO of a special issue collecting papers following the Euron GEM guidelines seen as akind of danish moviemaker Dogma89 for robotics

4. A collective book summarizing the contributions and the ideas coming from the GEM SIGand Euron benchmarking activities is being submitted to Springer STAR series.

The activities will continue after EURONII NoE end.

9/10


12/137


SIG Evaluation Criteria

We consider the GEM SIG successful as we were able to: define and document best practiceguidelines for robotics experimentation and benchmarking; define guidelines for GEM and

benchmarks of system performance in a number of core targeted areas; organize several specificworkshop on the matter at some major international conference ( IROS, RSS, ) The realambitious aim is to establish a research/publishing nucleus of activity adopting these principles andspreading the word to the whole international robotics community, with or without the publicationof JRRR. This is an ongoing, so far encouraging, activity.

A growing involved community of people sharing GEM ideas and spreading them has established.Some 20-30 people have directly contributed to the SIG activities so far. Including people not

previously involved in GEM SIG and or Benchmarking.

10/10


13/137

Activity ReportEURON Special Interest Group on Cooperative Robotics

Alessandro Saffiotti, Pedro Lima

April 30, 2008


14/137

1. Introduction

This document reports activities for 20072008 of the Special Interest Group (SIG) onCooperative Robotics of the European Robotic Research Network (EURON, http://www.euron.org).The purpose of this SIG is to meet the tremendously increased interest in cooperative robots ofdifferent types for many emerging applications, and to foster Europes position as leader in the field.This Interest Group was originally created in 2001 under EURON-1 to act as a catalyst inamalgamating the increasing number of European research groups in this field, and it has been activesince then.

On June 21, 2007, the SIG has been approved for sponsorship by the Board of EURON-2. The workto be performed by the SIG is described in the "Description of Work" document, which has beensubmitted to the EURON Board on September 25, 2007, and approved by the Board. The workplanned in that document spanned one year of activities, from October 1, 2007, till September 30,2008. The funded SIG activities have started as scheduled, and have initially developed according tothe plan.

Unfortunately, a major discrepancy in the planned activities was revealed at the beginning of 2008

due to the unexpected lack of authorization for extension of the EURON activities. All funded EURONactivities had to be completed by 30 April 2008, but many of the SIG activities were planned for afterthat date. This unexpected contingencies caused three changes in the SIG activities compared towhat had been planned in the "Description of Work":

One major activity has been anticipated to take place in April 2008;

Two major activities, which could not be rescheduled, will take place without the financial supportfrom EURON;

Because of this, not all of the allocated budget could be used.

The next section provides details on the activities that took place in the sponsored period (October 1,2007, to April 30, 2008), as well as on the originally planned activities which will take place later in the

year, although without the financial support from EURON. Section 3 summarizes the deliverables ofthe SIG.


15/137

2. Activities

The following list reports the originally planned activities, together with their status of achievement(including weather or not they could be supported by the EURON SIG)

Community portal

The community portal was created and put on-line on November 1, 2007, as scheduled. It comprisesa public web site, a private wiki space, and a mailing list. (Deliverable D1)

The website http://www.aass.oru.se/Agora/EuronCoop, which already existed from the previous SIGon Cooperative Robotics during EURON I (2001-2005), was re-designed, re-organized, andextended, with the ambition to become the main reference point for European researchers incooperative robotics. The site includes a description of the SIG objectives, regularly updated lists ofmember groups and of relevant events, and instructions on how to become members of the SIG, or tosubscribe to the SIG mailing list. The following figures shows the front page of the site.

Regarding the procedure to accept new groups as members of the SIG, any research group orcompany which is active in the field of cooperative robotics can apply for membership through theportal. Membership applications are decided by the SIG coordinators after an evaluation of the

relevance of the group research and development activities to the SIG objectives. In addition, anyinterested person can ask to subscribe to the SIG mailing list as an individual. Currently, the SIG has24 accepted members.

The mailing list is moderated in order to avoid spam, and it has been active but with low traffic untilnow. The list currently contains 55 members.

The Wiki space, accessible from the SIG web site, is intended as a space open for contribution fromall SIG members. Expected contributions to the wiki space include a student space to exchangeideas, references, and comments, and a catalog of relevant papers and software. Little activity hasoccurred until now on the Wiki: its use has been encouraged at the SIG meetings, and the SIG Boardmembers plan to take positive actions to put more contents of the Wiki in order to bootstrap theprocess.

This activity was achieved, with SIG logistic support.


16/137

Scientific events

The following scientific events were planned in the Description of Work.

Workshop on Network Robot Systems at IROS 2007

The workshop took place in San Diego, CA, USA, on October 29, 2007

(http://www.irc.atr.jp/iros07_nrs_workshop/), and it was co-organized by the SIG Board memberAlberto Sanfeliu. The workshop included nine paper presentations, plus a panel discussion ontestbeds and benchmarks for network robot systems. The proceedings of the workshop areattached to this document. (Deliverable D2)

This event was achieved, with SIG logistic support. Workshop on Formal Models and Methods for Multi-Robot Systems at AAMAS08

This workshop will take place in Lisbon, Portugal, 13 May 2008 (http://formal.isr.ist.utl.pt/),organized by the SIG Board members Pedro Lima and Nikos Vlassis, as well as by the SIGmembers Matthijs Spaan and Francisco Melo. The workshop will include 7 presentations, selectedby an International Program Committee, mostly composed of SIG Board members, and invitedtalks by Magnus Egerstedt (Georgia Institute of Technology, USA), and Alcherio Martinolli (EPFL,

Switzerland)

This event will take place, but without SIG financial support. To compensate for the missingsupport, the workshop will be mostly sponsored by one of the SIG member institutions, theInstitute for Systems and Robotics at Instituto Superior Tcnico, Lisbon

Workshop on Network Robot Systems at ICRA 2008

This workshop will take place in Pasadena, CA, USA, on May 19, 2008(http://www.irc.atr.jp/icra08_nrs_workshop/), and it is co-organized by the SIG Board memberAlberto Sanfeliu.

This event will take place, but without SIG financial support.

Educational events

A summer school on Cooperative Monitoring in Mixed Mode Environments will take place at TUDarmstadt, Germany, from 18 to 22 August 2008. Approximately 60 students are expected to attend.The SIG Board member Oskar von Stryk is in charge of the organization of this school.

This event will take place, but without SIG financial support. The main sponsor for the event will bethe Research Training Group on Cooperative, Adaptive and Responsive Monitoring in Mixed ModeEnvironments, funded by the German Research Foundation. The focus of the school has slightlyshifted because of this change in sponsorship.

SIG Meetings

A plenary SIG meeting took place as scheduled during the last EURON Annual Meeting, in Prague,Czech Republic, on March 28 2008. The meeting minutes are included in Appendix to this document.

This event was achieved, with SIG financial support.A SIG Atelier took place in Sarsted (close to Hannover), Germany, on April 20-21 2008. The goal ofthe Atelier was to prepare a White Paper on Cooperative Robotics in Europe, with special focus onthe two SIG topic themes for this year: Network Robot Systems and Formal Models and Methods forCooperation. (Deliverable D5). The Atelier was attended by nine SIG members: H. Levent Akin,Andreas Birk, Andrea Bonarini, Gerhard Kraetzschmar, Pedro Lima, Daniele Nardi, Monica Reggiani,Alessandro Saffiotti, and Matthijs Spaan.

This event was achieved, with SIG financial support.


17/137

3. List of deliverables

The following deliverables, planned in the SIG Description of Work, have been produced by this SIG(delivery date indicated in parentheses).

D1: Community portal (November 1, 2007)

D2: Proceedings of the NRS workshop at IROS-07 (November 30, 2007)

D5: White paper on European research on networked robotics and on formal models forcooperation (April 30, 2008)

D6: Final SIG Report (this document, April 30, 2008).

Deliverables D3 and D4 were scheduled for later this year.

Deliverables D2 and D5 are attached to this document. The minutes from the SIG meetings are inAppendix.


18/137

4. Financial report

The following table summarizes the expenses of the SIG for the reported period. Given that theoverall budgeted cost was of 15,000 euro, there is a surplus of about 5,000 euro.

EURON SIG on Cooperative Robotics - Expenses at 2008.04.30

Prague meeting

venue: 436

reimbursements: 2,941

A. Saffiotti 500

P. Lima 500

G. Kraetzschmar 500

V. Ziparo 480

E. Pagello 500

A. Pretto 461

Hannover meeting

venue: 1,464

reimboursements: 3,105

A. Saffiotti 800

P. Lima 432

G. Kraetzschmar -

D. Nardi 399

M. Reggiani 393

A. Bonarini 327

L. Akin 354

M. Spaan 400

Overhead (20%) 1,986

Total expenses 9,932

Received 15,000

Balance 5,068


19/137

Appendix: Reports on SIG Meetings

Plenary SIG Meeting in Prague, March 28 2008.

======================================================================EURON SIG on Cooperative Robotics -- Meeting of March 28, 2008

======================================================================

----------------------------------------------------------------------Agenda----------------------------------------------------------------------

1) SIG activity plan for 20082) Continuation of the SIG after EURON ends3) The SIG Wiki: how to bring it to full life?4) Planning for the SIG Atelier in Hannover5) Any other business

----------------------------------------------------------------------Minutes----------------------------------------------------------------------

Present 19 people.Chairing Pedro Lima and Alessandro Saffiotti.

0) Agenda agreement

- Agreed.

1) SIG activity plan for 2008

- Future activities were described to the SIG.

- The concept of "yearly themes" was reminded, to explain the emphasisin the SIG on "network robotics" and "formal models for cooperation"during 2008.

- It was pointed out that swarm robotics might be under-represented inthe SIG, some groups who are active in this field may join the SIGsoon.

- All people who are planning any related activity (eg, organization ofconferences, workshops, special issues, schools) please make sure tocommunicate them in due time to the SIG coordinators, so they can beincluded in the SIG web pages.

2) The SIG Wiki: how to bring it to full life?

- The current Wiki has been given some initial structure, but almostnobody has contributed to it.

- There was general agreement that this is a shared resources that wouldbe useful, and it should be used. It was felt that a bootstrap isneeded to get an initial critical mass, after that hopefully a "chainreaction" might begin.

- All SIG Board members will take an action to start populate the Wikipages, especially asking PhD students and young post-doc at their siteto do that.

3) Planning for the SIG Atelier in Hannover

- It was made clear that the main concrete objective will be to producethe "white paper on networked robot systems and on formal approachesto cooperative robotics", which is due as deliverable D5 of the SIG.

- It was felt, however, that we should try to also use this opportunityto discuss some further issues, which may be relevant to thecontinuation of the SIG (eg, future themes, future activities,possible uses of the "white paper", etc).

- Logistic information was provided.

- The coordinators will send a mail a couple of weeks before the Ateliersummarizing the logistic information, providing a backbone structurefor the white paper, proposing a schedule, and starting a discussionon the set of objectives of the Atelier.

4) Continuation of the SIG after EURON ends

- There was general agreement on having a yearly SIG meeting. Possible


20/137

means to organize this include: the Dagstuhl seminars in Germany, theFondation Treilles in France, the Centro Stefano Franscini at MonteVerita in Switzerland. One difficulty is that all these placestypically require to start the organization a couple of years before.

- The promotion of workshops co-located with major conferences shouldcontinue.

- Some people from the SIG are preparing a proposal for a Network of

Excellence on cooperative robotics, which might be continue and extendseveral of the SIG activities.

5) Any other business

- All SIG board members attending the meeting: please make sure to sendyour reimbursement documentation BEFORE APRIL 4 to the followingaddress:

Alessandro SaffiottiAASS Mobile Robotics LabDept of Technology, Orebro University70182 OrebroSweden

Send exactly the same form and document as you normally do for theEURON meeting. Please recall that: (a) you can claim expenses up to500 euro, (b) only documented expenses can be reimbursed, (c) youcannot get both reimbursement for the EURON meeting and for the SIGmeeting, (d) we will not be able to process late requests.

======================================================================


21/137

SIG Atelier in Starsted, April 20-21 2008.

======================================================================EURON SIG on Cooperative Robotics - Sarstedt Atelier, April 20-21 2008======================================================================

----------------------------------------------------------------------Agenda----------------------------------------------------------------------

April 20:9:00 agree on structure of document and schedule of this meeting9:30 brainstorm on the sections 2.2, 2.3 and 2.4, in that order12:00 assignment of sections to people12:30 lunch13:30 free time / homework on the assigned sections14:30 brainstorm on the sections 3.2, 3.3 and 3.4, in that order17:00 assignment of sections to people17:30 free time / home-work on the assigned sections

April 21:9:00 collect and compile together all assignments10:30 agree on how to proceed with the document

12:30 lunch13:30 future of the SIG15:00 end

----------------------------------------------------------------------Minutes----------------------------------------------------------------------

Participants: A. Bonarini, G. Kraetzschmar, A. Birk, M. Spaan, L. Akin,D. Nardi, M. Reggiani, P. Lima, A Saffiotti

The main part of the Atelier was devoted to the preparation of the WhitePaper, and work has proceeded following the agenda.

As for the part on the future of the SIG, the meeting started by revisingthe possible ways, raised in Prague, of proceeding with the SIG activitiesafter EURON ends:

- the best solution is if the NoE-ARC project gets approved

- There was general agreement on having a yearly SIG meeting. Possiblemeans to organize this include: the Dagstuhl seminars (applications:twice a year - 15/April, 1/November) in Germany, the Fondation Treillesseminas in France, the Centro Stefano Franscini seminars at Monte Veritain Switzerland, etc. Dagstuhl proposals approved fund the invitation ofabout 20 people.

- The promotion of workshops co-located with major conferences shouldcontinue (upcoming: SIMPAR 2008, Venice in November - 1 May 2008 is thedeadline for the proposals; DARS 2010 will be organized at EPFL; IROS2012 in Portugal)

- Enrico Pagelo suggested that we plan a special issue of ElseviersJournal of Autonomous Robotic Systems

The following actions were decided (by the SIG Coordinators, with the helpof the board members):

ACTION: check dates and conditions for the above seminar places, and forother alternatives: in particular, Pedro will check the possibility offinding a place and support to organize a meeting with NoE-ARC participantsin Lisbon, end of September/beginning of October 2008, and similarly for

Andreas Birk in Germany.

ACTION: if there's no funding from national organizations, we shouldprepare a workshop to be submitted at SIMPAR 2008.

In general, the opinion of participants was unanimous in that keepingthe SIG alive (with or w/o the EURON brand name) will strengthen futureproposals for project/NoE applications. This means for now, besides theabove actions, keeping the web site alive, and fostering the usage of theWiki.

======================================================================


22/137


23/137

Two "Hot Issues" in Cooperative Robotics:

Network Robot Systems, and

Formal Models and Methods for Cooperation

A white paper from the

EURON Special Interest Group on Cooperative Robotics

Date: April 30, 2008

Editors: Alessandro Saffiotti, Pedro Lima

Contributors: H. Levent Akin, Andreas Birk, Andrea Bonarini, Gerhard Kraetzschmar, Pedro Lima,

Daniele Nardi, Enrico Pagello, Monica Reggiani, Alessandro Saffiotti, Alberto Sanfeliu, Matthijs SpaanDissemination: Public


24/137

Executive Summary

This white paper on two currently hot areas on Cooperative Robotics research (Network Robot Sys-tems and Formal Models and Methods) aims to: (1) survey the state of the art for the two areas, (2)list in a justified manner their expected advances in the upcoming ten years, (3) identify theapplication topics of interest for Europe so as to keep its research competitive at the internationallevel, and (4) recommend lines of action for the support of EU research in Cooperative Robotics in thefuture. For each of the areas, this document discusses several key research aspects to achievecooperative intelligent behavior, spanning the current landscape of research in this field. Applicationswhere multi-robot systems interact in a natural way with humans at home, inside factories, and inhazardous environments are identified as the most promising domains where the EU may build abody of expertise which makes it the leader in the area of Cooperative Robotics. The intersection ofthe requirements for such applications with the expected research advances in future years and thecurrent expertise at the EU level provides a list of recommendations for research priorities for Europe.


25/137

Table of Contents

1. Introduction ...........................................................................................................41.1. Background: the Cooperative Robotics SIG............................................................................. 41.2. Purpose of this document......................................................................................................... 41.3. Structure of this document........................................................................................................ 5

2. Network Robot Systems.......................................................................................62.1. Definition of scope .................................................................................................................... 62.2. Current state of the art.............................................................................................................. 62.3. Expected advances .................................................................................................................. 7

3. Formal Models and Methods of Cooperation ................................................... 103.1. Definition of scope .................................................................................................................. 103.2. Current state of the art............................................................................................................ 103.3. Expected advances ................................................................................................................ 11

4. Conclusions.........................................................................................................13 4.1. The role of the two addressed areas in the field of cooperative robotics............................... 134.2. The role of European research in these two areas ................................................................ 134.3. Summary recommendations for EU research strategies........................................................ 14

References..................................................................................................................16


26/137

1. Introduction

1.1. Background: the Cooperative Robotics SIG

This document is a product of the Special Interest Group (SIG) on Cooperative Robotics of theEuropean Robotic Research Network (EURON, http://www.euron.org), funded by the EuropeanCommission. The purpose of this SIG is to meet the tremendously increased interest in cooperativerobots of different types for many emerging applications, and to foster Europes position as leader inthe field. This Interest Group, created in 2001 and active since then, intends to act as a catalyst inamalgamating the increasing number of European research groups in this field. During its sevenyears of activity, the SIG has provided an infrastructure for scientific exchange and high-level educa-tion in this area, and a point of contact between academia and industry. It has organized educationaland scientific events (e.g., Summer Schools, research ateliers, and scientific workshops at majorconferences) and maintained links to relevant international organizations and initiatives in the field,such as the IEEE RAS Technical Committee on Networked Robots, the RoboCup International Fed-eration, and the EURON Research Atelier on Network Robot Systems.

The key technology areas addressed by the SIG include: task allocation, cooperative planning andexecution, cooperative perception, multirobot mapping and localization, cooperative navigation, formalmodels of multirobot plans, multirobot learning, self-configuration, middleware for multirobot systems,truly heterogeneous cooperating robots, networked robotics, robot ecologies, and cooperation be-tween humans and multi-robot teams. An important horizontal issue is benchmarking of cooperativerobotic systems, including the definition of suitable benchmark scenarios and performance measures.The key application domains of interest to the SIG include: collaborative manipulation and trans-portation, space and underwater exploration, domestic robotics, entertainment, surveillance, searchand rescue.

Within the period 20072008, the SIG has been supported by the EURON FP6 Network of Excel-lence for a period of 8 months. For this period of activity, two main areas were identified as its maintopical foci, due to their comprehensiveness, their complementary nature, and the current interest of

European researchers in the SIG: Network Robot Systems and Formal Models and Methods for Co-operation. This white paper on the status of European research concerning these areas constitutesone of the main deliverables of this SIG for this supported period, as specified in the SIG Descriptionof Work (available at http://aass.oru.se/Agora/EuronCoop/ ). Other deliverables for this pe-riod include an Internet community portal (see previous link), proceedings of workshops organized atthe IROS-2007 and AAMAS-2008 conferences, and the organization of a Summer School to be heldin August 2008 in Darmstadt, Germany.

The present document was conceived during a SIG atelier organized for this purpose on April 20thto 21st, 2008, in Sarstedt, Germany, where the contributors met to thoroughly discuss the above top-ics. The final version of this document reports the results of those discussions, and it has been com-pleted and refined by all the participants in the week following that meeting.1.2. Purpose of this document

The purpose of this document is threefold:

1) to outline the current status of research on the two focal areas of interest (Network Robot Sys-tems and Formal Models and Methods for Cooperation) in Europe and worldwide;

2) to forecast the plausible evolution of the research around these areas, both in terms of researchpush and application pull;

3) to provide recommendations and priorities for European research strategies in order to keep aleading role within this field.

For each of the two focal areas, the white paper covers several research aspects as guidelines for the

above purposes.


27/137

1.3. Structure of this document

This document is organized as follows. Section 2 deals with Network Robot Systems, starting with adefinition of the intended scope of this term, and continuing with a discussion of the current state ofthe art and of the expected advances for research in this area, respectively. Section 3 deals with For-mal Models and Methods for Cooperation, using the same structure as Section 2. Section 4 summa-

rizes the recommendations for research in the EU regarding those areas.


28/137

2. Network Robot Systems

2.1. Definition of scope

The term "Network Robot System" (NRS) comes from the Network Robot Forum, a mixed industry-academia platform with over 100 members in Japan, whose aim is to establish basic technologies toallow different robots distributed in the environment to work collectively. This forum extended the defi-nition of a robot to also include software agents and static sensors in the environment. In Europe, thisterm was adopted by the EURON-II Research Atelier on Network Robot Systems [SA08]. NetworkRobot Systems are also within the scope of the IEEE RAS (Robotics and Automation Society) Tech-nical Committee on Networked Robots (http://faculty.cs.tamu.edu/dzsong/tc). The latter originally fo-cused on Internet-based teleoperated robots, but it expanded in 2004 to cover autonomous systemswhere robots and sensors exchange data via the network.

For the purpose of this document, we consider a Network Robot System to be any distributed sys-tem which consists of a multitude of networked robots and other devices and which, as a whole, iscapable of interacting with the environment through the use of perception and action for the perform-

ance of tasks.

2.2. Current state of the art

Network Robot Systems is a relatively young research field, still quite dynamic and open to ideasand concepts coming from other research areas.

Initial activities were bootstrapped by the advance in network infrastructure and internet programming.Ten years ago, the reduction of time delays in communications and the improvement in architecturalrobustness allowed the development of internet-based robotic teleoperation. Several online robotapplications were developed [GS02] allowing users from all over the world to control remote robotichardware to paint [Ste00], excavate [GGS+00], explore [Sim98], and exit from mazes [SM00].

While teleoperated robots are still valuable resources, the research attention has recently opened to abroader set of problems and applications. To reflect this change, in May 2004, the related IEEE RASTechnical Committee changed its name from Internet and Online Robots to Networked Robots. Justbefore, in 2003, the Japanese Council for Science and Technology Policy supported the creation ofthe Network Robot Forum (NRF, http://www.scat.or.jp/nrf/English). Currently, this forum involves overa hundred members from industry, academia, and government.

A new stream of research started as an extension of the concept of sensor networks. The basic ideais to introduce mobility either adapting the geographical distribution of sensors based on the acquiredinformation [HPS04, Suk] or supporting the deployment of autonomous robotic systems that move inthe field [BSH04, MVO05, EG06, OSCS07]. Significant projects include situational awareness usingsensor networks [KRS04, KKY+06], environmental robotics to observe, monitor, and assess the stateof complex environmental processes [PSK+04, ZSR04, BRY+04], and monitoring, deployment, and

repair of sensor networks [CHP+04].An additional class of networked robot applications has been recently introduced based on the re-search on ubiquitous computing [Wei91, Sat02, ECPS02]. This research is a major trend in informa-tion technology resulting already in home appliances containing sensors and becoming networked.The development toward ubiquitous robotics resulted in a natural step to integrate networked robotswith ubiquitous computing environments that include networked sensors and actuators as well as hu-man beings [KKL04]. The concept of ubiquitous robotics was further expanded by a Japanese studygroup on network robot organized by Ministry of Internal Affairs and Communication of Japan to in-clude three types of network robots: visible, virtual, and unconscious [AH06, LOH06, TM07]. The visi-ble-type of robots embodies the traditional concept of robots, such as humanoid or industrial devices.Virtual robots are software agents that exist only in a virtual space and interact with users throughPCs and mobile phone displays. Finally, so-called unconscious robots identify robots of which the

users is not aware but that are required to collect information for the actuation of visible and virtual ro-


29/137

bots. This class mainly refers to smart sensors embedded in roads, rooms or equipments, or hidden inclothes or accessories.

Within the concept of ubiquitous robotics, a few joint EU projects have recently started their activitiesin Europe. The URUS project (FP6-IST-045062, http://urus.upc.es/) aims at developing a cognitivenetwork architecture integrating robots, environment sensors and humans, all connected through Wi-

Fi technology. The main scientific challenges are the development of key tools for NRS, including co-operative navigation and localization, cooperative perception, cooperative map building, and task al-location, while the main application areas include surveillance in urban areas, transportation of goods,and people guidance and assistance [SA06]. Another European project, DUSTBOT (FP6-IST-045299,http://www.dustbot.org), aims at creating a system for garbage collection and air quality monitoring inurban environment. System functionalities are provided through the integration of wireless sensornetworks and mobile robots. Finally, the AWARE project (FP6-IST-33579, http://www.aware-pro-ject.net) aims at the development of technologies for the cooperation of flying platforms and groundsensor networks. One of the main challenges here is the development of an architecture and a mid-dleware supporting cooperation among the systems.

At the national scale in Europe, a relevant project is the PEIS-Ecology (http://aass.oru.se/~peis) inSweden, which was initiated as a collaborative effort between rebro University in Sweden and ETRI

(Electronic and Telecommunication Research Institute) in Korea. This project is distinct in its empha-sis on the fundamental scientific principles that underlie the design and operation of an ubiquitous ro-botic system, such as middleware for highly heterogeneous distributed systems, self-configurationand dynamic re-configuration, cooperative perception, and the integration between digital and physi-cal interaction [SB05, SBSC07]. In Italy, the national project APE (Agents for Perception in Environ-mental Monitoring) put in evidence how cooperative robots can actively participate to reconfigure adynamical sensor network (DSN) for monitoring pollutants in the environment [ABC06].

Outside Europe, the largest attention to the development of ubiquitous robotics comes from Japanand Korea. In Japan, four major players (NTT, Toshiba, Mitsubishi Heavy Industries, and ATR) startedthe Network Robot Project (http://www.irc.atr.jp/ptNetworkRobot) in 2004, to improve people under-standing of their situation and environment through the development of a Network Robot System.Four main aspects are addressed: communication among robots, the networking platform, ubiquitous

sensing, and human-robot interaction. This project has a strong focus on the human-robot interactionin real environment. Two main field experiments in real world have been already carried out to high-light intrinsic research issues. A first one involved a Japanese elementary school [KHEI04] to analyzethe relationship between pupils and robots. A second one was at the Osaka Science Museum toguide visitors and to motivate them to study science [KSP+06].

In Korea, the URC project introduced the concept of Ubiquitous Robotic Companion (URC) [Oh03] asa vision where a ubiquitous service robot provides users with the services they need anytime, any-where in ubiquitous computing environments [HSCY05]. The emphasis of this project is on reducingthe complexity of physical robots by delegating computationally intensive tasks to an external server.Field tests of various scales had been conducted to verify the effectiveness of the approach in taskssuch as monitoring and security.

2.3. Expected advances

Research in Network Robot Systems has constantly grown over the past decade. Europe holds astrong position in this field, and it could become a world leader in the near future if enough supportis provided. Provided that positive measures will be taken in order to sustain and strengthen this posi-tion, the European research community in Network Robot Systems will be able to produce major ad-vances and breakthroughs in the field.

The technological foundation of these advances will be based on past and current progress in thecomponent technologies and in research platforms. As research in Network Robot Systems requirestest environments featuring a multitude of autonomous robots, smart sensor devices, and other actu-ated or sensor-enabled components, the free availability of and access to suitable test-bedsanddevices at low or reasonable cost would significantly lower the entrance barrier to this area, extend-ing the possibility of its applications. Within the next three years, we expect more reliable, robust,flexible, smaller, and cheaper autonomous mobile robots to appear on the market, for both indoor and


30/137

outdoor applications, while market forces in the safety and security area are likely to produce abroader range of components for embedded and wireless sensor networks, such as intelligent cam-eras, smart sensors, and devices for bio-identification. Essential for making this technology availableand usable for research on network robot systems will be the development of suitable middleware,which helps to manage the heterogeneity of the underlying hardware infrastructure and provides net-work-transparent access to the services of the system components, and the integration into an appro-

priate tool chain.Based on the availability of an improved research infrastructure, we expect the following ad-vances in the next ten years:

High-level cooperative cognitive skills: while there is a substantial need for improvement ofindividual cognitive skills, the ability to achieve cooperation in planning, decision making and envi-ronment modeling is key to the development of NRS.

Autonomy: Networked robots must be able to accomplish tasks in a largely unsupervised way,typically teaming with other robots in order to improve over the individual capabilities and ensuresystem robustness. Moreover different degrees of autonomy, as well as trust have a key role innetworked robots, where such notions must suitably accommodate the heterogeneity of the ro-bots in the network.

Modeling: Natural cognitive system have great capabilities of developing behaviors within a net-work. While the concept of network in natural systems has typically been regarded within a com-munity of homogeneous (and often cognitively very simple) agents, natural networks can be muchmore complex including heterogeneous agents, as well as opponents.

Behavior representation and interpretation: In a network of robots both the representation ofthe environment and its interpretation are performed cooperatively and in a distributed fashionwithin the network. Moreover, learning approaches can be developed to learn collectively andlearn collective behaviors.

Human/robot interaction: Better interfaces to control and interact with NRS will improve usabilityand make new, broader applications possible. On the one hand, improved distributed cooperativeperception capabilities of NRS will make it possible to have effective interaction with people, byunderstanding different kinds of signals coming from single and multiple persons sharing the NRSspace; on the other hand, a scenario with multiple users interacting with multiple robots bringsabout new challenges that will significantly impact on human/robot interaction.

Performance Evaluation and Benchmarking: Suitable methods must be established in order tosuitably assess the features of the proposed methods. This requires both a methodology for de-veloping effective experimental validation procedures, shared experimental settings (both insimulation and on real robots) and benchmarks. The RoboCup community offers a great exampleof a well assessed benchmark for cooperative skills of multi-robot systems, although achieved insomewhat ad hoc environment

It should be emphasized that many of the above advances naturally extend to NRS the challenges forcognitive systems highlighted in the FP7 programme. This has to be expected, since extending the

horizon of the current research on cognitive and robotic systems from single-robot systems to multi-robot systems is a natural step that will enable the development of systems with higher degrees offlexibility, adaptivity, and robustness.

The technological advances on the above topics will enable the development of new applications ofNRS with a strong impact on economy and society. Foremost of all, such applications are expected inthe wide area of safety and security, in environments encompassing personal homes, public build-ings, and wider public spaces.

Among the applications of NRS which are regarded as viable in the next ten years we mention:

Elderly care, where a NRS can assist elderly people in their personal homes or retirementhomes, with the goal to provide physical and cognitive support, to facilitate communication withand monitoring by remote relatives and care givers, and to detect and respond to emergencies.


31/137

Security applications and intelligent buildings, where a NRS can track and classify the behav-ior of people, hazardous situations or threatening acts can be detected, and actions can be de-cided to maintain a safe status.

Networked service robot systems, for applications such as trash collection, delivery and logis-tics, both in public spaces (e.g., city streets) and in private workplaces (e.g., factory floors).

Flexible automation and collaborative manufacturing, where heterogeneous, cooperative ro-botic systems are expected to set forth the future of networked control in industrial settings.

Wide-scale environmental monitoring, by deploying autonomous sensor networks with the abil-ity to self-deploy, self-reconfigure, and self-repair, capable of monitoring in a largely unsupervisedway large environments for pollutions, environmental theats, and other hazardous situations.

Cooperative search and rescue, where NRS actively search for people or objects, and eventu-ally support rescue personnel in dealing with emergencies.


32/137

3. Formal Models and Methods of Cooperation

3.1. Definition of scope

Formal models and methods of cooperative robot systems aim at providing modeling, analysis,benchmarking, learning, and design-from-specifications tools for problems pertaining to such systems(e.g., motion coordination, task planning, mapping). Examples of formal models and methods for co-operative robots include, but are not limited to, logic-based knowledge representation and planning,decentralized decision-making, graph-based control methods, game theory, Bayesian networks, dis-crete and hybrid system models, or models of natural systems applicable to robotics. Examples oftools include plan verification, modeling a robotic task as a discrete event system, or quantifying theuncertainty level of a localization method.

For the purpose of this document, the characterization of formal models and methods for coopera-tive robot systems is made by considering the level of abstraction and generalization attemptedand achieved by a given approach: a necessary requirement for a formal method is that the resultingmodels and methods apply to a more general class where that problem belongs to, providing a sys-

tematic approach to problems in that class. Ad-hoc approaches, whose goal is to solve a specificproblem in a way that can not be generalized to other similar problems, can not be considered as pro-viding formal models and methods.

3.2. Current state of the art

Formal models and techniques have been developed to build successful cooperative multi-robotsystems, providing solutions for several types of problems [LVSM08].

An inherent property of a multi-robot team is the fact that sensors are spatially distributed, and appro-priate techniques have been developed for sharing and fusing the information coming from distrib-uted, heterogeneous sources. The goal of sharing sensor information via communication is to raisethe level of situational awareness, allowing for better task performance. Localization is, for robots, abasic requirement, when it comes to cooperative spatial perception. The localization problem ismeanwhile predominantly addressed together with mapping, which leads to the paradigm of Simulta-neous Localization and Mapping (SLAM). Most current approaches to SLAM are based on probabilis-tic (Bayesian) approaches, typically employing Kalman filter methods [DNC+01], particle filters or ex-pectation maximization techniques [TBF05]. While SLAM is well established for 2D mapping by singleland robots, multi-robot mapping is considered to be a very important but also still largely open prob-lem [KFL+03, FKK+06]. For cooperative perception of features of the environment, distributedBayesian sensor fusion techniques have been developed [MCF+07]. In this context, a fusion ruleshould take into account that the same information may reach a robot several times due to loops inthe information channels [GD94]. In addition to these quantitative approaches, qualitative and logic-based representations have also been applied, including some based on multi-agent belief logics[KH03], on fuzzy logic [HRWI07], and on conceptual spaces [LS08].

Formal models for multi-robot plans provide a significant step in defining suitable solutions forcooperation. Cooperation in multi-robot systems plays an important role, as teamwork can lead toconsistent performance improvements. Several approaches achieve cooperation facilitatinginteraction through the assignment of individual behaviors [FINZ06, DS02, INPS03, Par98, GM00,CCK02] or through the automatic generation of cooperation patterns [LKS07]. A formal analysis oftask allocation approaches for multi-robot systems has been studied [GM04]. Some works havestudied the possibility of a structured approach to the design of cooperation, for which coordinationand synchronization is required. In several works, e.g., [YOW+98, PDM+99], the engagement in acooperative behavior is usually not explicitly modeled, and it is difficult to handle situations, such asaction failures, in which the robots have to withdraw the cooperative execution.

The complexity of behavior specification in real domains requires formal tools for plan validation

which, generally, cannot be provided by ad-hoc solutions. Petri Nets are an appealing modeling toolfor Discrete Events Systems, which has been used in several works for the modeling of robotic be-haviors, and which provides the means for formal validation of important properties such as reachabil-


33/137

ity or deadlocks. In a cooperative robotics context, Petri nets have been used for modeling multi-robotplans [CL08, ZIN+08], and a multi-robot coordination algorithm for environment exploration [SY05].Recent work explores combining Linear Temporal Logic and Discrete Event Systems SupervisoryTheory to go from complex task specifications to robot task controllers [KFP07]. Furthermore, alsomulti-agent Belief-Desire-Intention logics can be used for plan verification [BFVW06]. Decentralizedpartially observable Markov decision processes (Dec-POMDPs) form a rich mathematical framework

for representing cooperative planning under uncertainty problems [SS08, OSV08]. For instance,[EMGST05] demonstrated the viability of approximate Dec-POMDP techniques for controlling a smallgroup of robots, applying cooperative game-theoretic concepts. Game theory also provides tech-niques to plan for multi-robot teams in the presence of adversaries. Finally, also formal approaches tomulti-robot formation control have been proposed (e.g., [SEH08] for a recent sample from a large setof references).

Learning is a core issue for cooperative robotics. First, learning and adaptation are essential to cre-ate multi-robot systems that are robust, scalable and most importantly that produce increasedbenefits. Second, multi-robot systems are affected by particular challenges due to their distributednature and multiple goals, as well as noise in sensing and acting, complicating a straightforward ap-plication of standard machine learning techniques. For instance, many convergence proofs for rein-forcement learning depend on a stationary environment, i.e., with only one learning robot [FP01,

GY04]. In [WTS07], evolutionary and temporal-difference approaches to reinforcement learning arecompared in a robotic soccer context. Learning has been successfully applied in behavior-basedsystems to adapt task assignment in heterogeneous robot teams, dealing with individual robot capa-bilities that change over time [Par00]. Unsupervised learning methods such as evolutionary algo-rithms have been popular for multi-robot task optimization [MC96, PM06]. Further opportunities forlearning in cooperative robotics concern adaptive navigation and exploration, and opponent modeling[BM00].

3.3. Expected advances

A key general issue in cooperative multirobot systems, and network robot systems in particular, is thedevelopment of new models and cooperation paradigms to deal with the complexity involved in

these systems. Furthermore, traditional solutions are centralized and require high levels of connec-tivity, impose a substantial computational burden, and are typically more sensitive to failures andmodeling errors than decentralized schemes. In recent years, distributed solutions for perception,planning and control have been proposed. However, in general, there is a lack of methodologies inrobotics dealing with the formalisms needed for the analysis and design of multiple robots interactingin real time among them and with the infrastructure embedded in the environment. This need is par-ticularly evident when considering heterogeneous robots and environment sensors, robots interactingsimultaneously with multiple objects and humans in the environment, scalability and quality of servicefor a large number of robots and interacting objects.

Heterogeneity and interaction with humans increases complexity, but can also be a source ofbenefits when considering the exploitation of the complementarities of the robots for perception andactuation. However, formal methods to address the optimal real-time cooperation of heterogeneousrobots and humans, exploiting their complementarities, are also needed.

Many formal methods can only be applied at a reasonable computational cost with a small number ofrobots. Furthermore, the number of devices (sensors and actuators) and humans interacting in realtime with the robots is an important issue: currently, the existing communication infrastructure enablesthe interaction of hundred or even thousands of humans with suitable communication devices such asmobile phones, PDAs or laptops. Pressed by those needs, formal models that enable scalability arecertainly expected to develop in upcoming years.

Therefore, in the near future we foresee the need for a substantial improvement of formal methods forcooperation, with the aim of better understanding the underlying theoretical principles, ensuringsafety, robustness and adaptation, and/or for developing more systematic or automated methods forthe design and implementation of cooperative robotic systems, in the following key areas:

Cooperative planning and execution: The added value brought by formal methods extendsnaturally to domains in which multiple robots plan and execute coordinated plans, smoothly inter-


34/137

acting to cooperate in the achievement of individual and collective goals. A characterizing featureof multi-robot domains is the uncertainty arising from both perception and action, which requiresthe introduction of sensing actions. Another significant modeling requirement is brought by thepresence of other agents that can interfere, while pursuing their own goals or even by havingcompeting goals. In this context, the research endeavor is expected to focus on specificationlanguages for multi-robot plans, plan analysis, integration of action models and process models,

robot team strategies for task planning and decision making, distributed/decentralized plansynthesis and execution under uncertainty, combined coordination and cooperation in dynamic,uncertain, competitive scenarios, and methods for dynamic team formation and task assignmentunder uncertainty

Cooperative perception: Here, the goal is to achieve systems that, in a distributed way, arecapable of gathering and interpreting sensor data from the environment, leading to both individualand collective understanding of the situation that is functional to intelligent behavior. Use of formalapproaches is expected to lead to an increase in the ability of providing formal guarantees interms of safety, dependability and fault tolerance. A significant number of tasks deeply rely on co-operative perception, e.g., cooperative situation assessment, cooperative tracking or cooperativelearning from sensor information. Three main problems in cooperative perception that warrantdeeper theoretical analysis are information fusion, trust, and uncertainty. Information fusion is al-

ready well-studied for the cases of uni-modal, multiple-source and multi-modal sensor fusion,where the information sources (i.e., sensors, either mobile or not) are all known from the begin-ning and trusted. New formal models are needed especially when the set of sources varies overtime (more or less rapidly) and cannot be implicitly trusted, and when one needs to fuse informa-tion from sources for which there is no error model known, or sources that keep changing theirspatial location over time.

Cooperative learning: The goal is to achieve, most likely through a layered approach, the abilityfor a team of robots to learn the features of the environment and, when relevant, opponent mod-els. Moreover, a team of agents should be able to learn collective behaviors, such as strategies topursue their goals in the environment, in the face of competitors. Learning in multirobot systems isaffected by specific challenges like multiple goals, noisy perception and actions, and inconsisten-cies in the internal states and in environment models between the individual robots. A straightfor-ward application of standard machine learning techniques is therefore difficult and even ineffec-tive in some cases, which require the development of specialized methods. There are severalmajor open problems in multi-robot learning including modeling formal properties of real worlds,convergence time of learning algorithms, and coping with dynamical environments including otherrobots learning.

Evaluation: The goal is to come up with metrics and benchmarks that enable a systematicperformance evaluation of the proposed models and methods. The metrics can also be used inreinforcement learning algorithms for defining rewards.

Finally, significant progress will not be achieved in the next ten years without completely new ap-proaches to modeling and implementing cooperation that go beyond the current horizon. Indeed, theability to build models and methods to design robots that can work cooperatively is the key to the suc-

cessful deployment of robots in the society.


35/137

4. Conclusions

4.1. The role of the two addressed areas in the field of cooperative

roboticsMulti-robot cooperative systems extend the capabilities of single robots to carry out tasks requiringcooperative skills. In recent years, ubiquitous networked devices are becoming common place,therefore any autonomous robot should be prepared to interact and cooperate with other devices inthe environment. Moreover, several simpler robotic devices can be combined into a distributed systemto achieve levels of competence, flexibility and robustness beyond what can be achieved today bya standalone robot or multiple traditional robots; in particular, the inherent distributed nature of such asystem can provide a level of fault-tolerance that permits graceful degradation of service availability ifsingle subsystems fail. Finally, a distributed system offers obvious practical advantages to the enduser in terms of modularity, configurability and extensibility. Generalizing the concept of multiple ro-bots to a set of devices which include sensors (some of them mobile, since they are assembled onrobots) and actuators, networked by a communication system, possibly wireless, effectively merges

the multi-robot cooperative systems concept with other concepts, namely those of sensor networksand ambient intelligence, yielding what is known as Network Robot Systems. NRS design is chal-lenging due to its complexity and large-scale nature. Therefore, methods that provide a systematicapproach to NRS design, ensuring robustness, flexibility, scalability, and effectiveness, are desirableas well. While such issues are addressed by the design of swarms, composed by individuals typicallyhomogeneous and with simple capabilities but collectively capable of interesting behavior, NRS spe-cifically aim at more complex systems, where the agents are highly heterogeneous, can build andmaintain high-level cognitive models of the environment, and can flexibly perform a variety of tasks,possibly through sophisticated interaction with human users.

The two areas covered in this white paper address the above endeavors and are complementary innature. Network Robot Systems mainly concerns the technology involved in applications consistingof several sensors and actuators connected by a communication network,. Such applications have a

scope large enough to encompass most current cooperative robotics applications, and are expectedto have a high social and economic impact. Formal Models and Methods focus on theory andmethodologies to devise novel and well-founded solutions for cooperative robotics, given availabilityof an infrastructure technology, namely a network robot system.

4.2. The role of European research in these two areas

Application scenarios have evolved over the past years from specific, single-robot systems to ones inwhich multiple actors, both artificial and human, cooperate towards the achievement of objectives. Asa consequence, the capability of teams of robots to coordinate their activities is acquiring great strate-gic importance for the inclusion of robotic technologies in increasingly many domains of activities. Fu-ture research needs to invest massively in theories, methods and applications of cooperative robot-

ics if it is to facilitate the uptake of fundamental research results in robotics.The ability of multi-robot systems to coordinate their activities requires high-level cognitive skills, in-cluding the ability to plan, negotiate, achieve meaningful communication, adapt to changes in the be-havior of the other actors or of the environment, and dealing with failures. Europe holds a worldleading position for its research on artificial cognition, and it can therefore become a worldleader in research on cooperative robotics.

In the reverse direction, progress in multi-robot system will be pivotal to the further progress ofEuropean research on artificial cognitive systems, and to their pervasive inclusion in the Euro-pean society. In fact, extending the horizon of the current research on robotic systems from single-ro-bot systems to multi-robot systems is a natural step that will enable the development of systems withhigher degrees of flexibility, adaptivity, and robustness. In the long run, this research will enable thedevelopment of cognitive symbiotic systems that include both robots, humans, and software agents.


36/137

The two research areas discussed in this paper, Network Robot Systems and Formal Models andMethods for Cooperation, are expected to be critical enablers for research in cooperative robotics.Europe is ideally placed to become the world leader in these two research areas , because of itsrole as a pioneer in basic and applied research in Ambient Intelligence, and because of its undis-puted position in the development of formal theories and methods to be used as a foundation of noveltechnologies.

Cooperative robotic systems have strong application in domains which are of strategic impor-tance to European social and economic priorities. Systems of cooperating robots and other de-vices interacting with humans can be deployed in domestic environments, to provide increased safety,assistance and comfort to citizens. In particular, systems of this type will be paramount in elder-careassistance to improve the safety, independence and quality of life of senior citizens, thus facilitatingaging in place. Systems of this type can also be deployed in the environment to provide environ-mental monitoring to help in the prevention, detection and intervention in case of environmental andsecurity crises. They can be deployed in public and private places to provide standard robotic serviceslike transportation and cleaning with greater flexibility and reliability. Finally, the current high levels ofinvestment in networked control systems has demonstrated that certain segments of the industry arestrongly inclined to take up the products of research in cooperative robotics to build systems able toprovide modular solutions to flexible automation and collaborative manufacturing. The development of

robust solutions in any one of the above application domains is an important social and/or economicobjective for Europe, as set up in the FP7 agenda.

4.3. Summary recommendations for EU research strategies

While European researchers have produced results at the forefront of international research in thefield of cooperative robotics, this field has not been explicitly prominent in the EU research strategiesuntil recently. The above discussion shows that this field should now be considered as a majorresearch priority for the development of future cognitive robotic systems. In particular, this white pa-per has identified the areas of Network Robot Systems and of Formal Models and Methods for coop-eration has two areas which:

1) are critical enablers for future breakthroughs in the field of cooperative robotics that are both well-founded and relevant to potential applications of relevance to EU policies; and

2) constitute an adequate niche for EU research, where the strong competences of Europe can beexploited.

Within these areas, a clear intersection with the FP7 programme has been highlighted, while at thesame time realizing that there is currently no specific focus in that programme on the ability to developcooperative/distributed approaches to many of the key issues that are therein specified. Without aspecific focus on the cooperative aspects discussed in this paper, the overall goal of the programmemight lack one key enabling element. Specifically, we recommend to address:

a) High-level cooperative cognitive skills

b) Autonomy to enable cooperation

c) Modeling cooperation in natural cognitive systems

d) Cooperative Behavior representation / interpretation

e) Human/Robot-teams Interaction

f) Performance Evaluation and Benchmarking for cooperative systems

For each of these issues, solid results already exist in the case of single-robot case, but a majorqualitative leap is needed in order to extend those results to the case of cooperative robot systems.In addition, we expect that the following topics concerning Formal Models and Methods for Coopera-tion deserve special attention:

g) Cooperative planning and cooperative execution, also in presence of humans;

h) Cooperative perception, also including the ability to interact and to share information with hu-mans;


37/137

i) Learning and adaptation in multi-robot, cooperative systems.

Within each of the above issues, research should aim at the development of sound theories andmethods, where "soundness" is defined with respect to:

Development and exploitation of formal models and methods;

Evaluation, based on sound methodologies and using standardized benchmarks and metrics.The last point in particular points to the need to develop common platforms, test-beds and bench-marks specifically geared toward the evaluation of cooperative robotic systems. A European effort inthis direction would also facilitate the sharing of knowledge and results among researchers in Europe,thus acting as a critical catalyst to further progress in this field.


38/137

References

[ABC06] F. Amigoni, A. Brandolini, V. Caglioti, V. Di Lecce, A. Guerriero, M. Lazzaroni, F. Lombardi, R. Ottoboni,E. Pasero, EV. Piuri, O. Scotti, and D. Somenzi, "Agencies for Perception in Environmental Monitoring".IEEE Transactions on Instrumentation and Measurements, 55(4):1038-1050, 2006.

[AH06] T. Akimoto, N. Hagita, "Introduction to a Network Robot System", Int. Symposium on Intelligent SignalProcessing and Communication Systems (ISPAC), 2006.

[BFVW06] R. H. Bordini, M. Fisher, W. Visser, and M. Wooldridge, "Verifying Multi-agent Programs by ModelChecking". Journal of Autonomous Agents and Multi-Agent Systems12(2):239-265, 2006.

[BM00] A. Bonarini and M. Matteucci, "Learning context motivation in coordinated behaviors". Proceedings of theSixth Intelligent Autonomous Systems Conference (IAS-6), IOS Press, Amsterdam, NL, 519-526, 2000.

[BRY+04] M. Batalin, M. Rahimi, Y. Yu, S. Liu, G. Sukhatme, W. Kaiser, M. Hansen, G. Pottie and E. Estrin, "Calland Response: Experiments in sampling the environment", ACM SenSys, 2004.

[BSH04] M. A. Batalin, G. S. Sukhatme, and M. Hatting "Mobile Robot Navigation using a Sensor Network", Int.Conf. on Robotics and Automation (ICRA), 2004.

[CCK02] L. Chaimowicz, M. F. M. Campos, and V Kumar, "Dynamic role assignment for cooperative robots". InProc. of the 2002 IEEE Int. Conf. on Robotics and Automation (ICRA02), pages 292 298, Washington DC,May 2002

[CHP+04] P. I. Corke, S. E. Hrabar, R. Peterson, D. Rus, S. Saripalli, and G. Sukhatme, "Deployment andConnectivity Repair of a Sensor Net with a Flying Robot". Int. Symposium on Experimental Robotics(ISER), 2004.

[CL08] H. Costelha and P. Lima, "Modelling, Analysis and Execution of Multi-Robot Tasks using Petri Nets".Proc. of 7th International Joint Conference on Autonomous Agents and Multi-Agent Systems, Estoril,Portugal, 2008

[DNC+01] G. Dissanayake, P. Newman, S. Clark, H.F. Durrant-Whyte, and M. Csorba, "A solution to thesimultaneous localisation and map building (SLAM) problem". IEEE Transactions of Robotics andAutomation, 17(3):229241, 2001.

[DS02] M. D. Dias and A. Stentz, "Opportunistic optimization for market-based multirobot control". In Proc. ofIEEE/RSJ Int. Conference on Intelligent Robots and Systems (IROS02), pages 27142720, Sept. 2002.

[ECPS02] D. Estrin, D. Culler, K. Pister, and G. Sukhatme, "Connecting the Physical World with PervasiveNetworks", IEEE Pervasive Computing, 1(1):59-69, 2002.

[EG06] I. H. Elhajj and J. Gorski, "Sensor Network and Robot Interaction using Coarse Localization", IEEE/RSJInt. Conference on Intelligent Robots and Systems (IROS), 2006.

[EMGST05] R. Emery-Montemerlo, G. Gordon, J. Schneider, and S. Thrun, "Game theoretic control for robot

teams". In Proceedings of the IEEE International Conference on Robotics and Automation, 2005.[FINZ06] A. Farinelli, L. Iocchi, D. Nardi, and V. A. Ziparo, "Assignment of dynamically perceived tasks by token

passing in multi-robot systems". Proceedings of the IEEE, Special issue on Multi-Robot Systems,94(7):12711288, 2006.

[FKK+06] D. Fox, J. Ko, K. Konolige, B. Limketkai, D. Schulz, and B. Stewart, "Distributed Multirobot Explorationand Mapping". Proceedings of the IEEE, 94(7):1325-1339, 2006.

[FP01] F. Fernandez and L. E. Parker, "Learning in Large Cooperative Multi-Robot Domains". InternationalJournal of Robotics and Automation, special issue on Computational Intelligence Techniques inCooperative Robots, 16 (4):217-226, 2001.

[GD04] S. Grime and H. F. Durrant-Whyte, "Data fusion in decentralized sensor networks". Control EngineeringPractice, 2(5):849863, 1994.


39/137

[GGS+00] K. Goldberg, S. Gentner, C. Sutter et al., "The Mercury Projects: a Feasibility Study for InternetRobotics", IEEE Robotics and Automation Magazine, 7(1):35-40, 2000.

[GM00] B. Gerkey and M. J. Mataric, "Principled communication for dynamic multi-robot task allocation". InProceedings of the Int. Symposium on Experimental Robotics, pages 353362, Waikiki, Hawaii, 2000.

[GM04] B. Gerkey and M. J. Mataric, "A formal analysis and taxonomy of task allocation in multi-robot systems".

Intl. Journal of Robotics Research, 23(9):939-954, September 2004.

[GS02] K. Goldberg, R. Siegwart, "Beyond Webcams: an Introduction to Online Robots", MIT Press, 2002.

[GY04] D. Gu and E. Yang, "Multiagent Reinforcement Learning for Multi-Robot Systems: A Survey". Technicalreport CSM-404, Department of Computer Science, University of Essex.

[HLNY07] K. Han, J. Lee, S. Na, and W. You, "An Ambient Robot System based on Sensor Network: Conceptand Contents of Ubiquitous Robotic Space", Int. Conf. on Mobile Ubiquitous Computing, Systems, Services,and Technologies, 2007.

[HPS04] A. Howard, L. E. Parker, and G. S. Sukhatme, "The SDR Experience: Experiments with a Large-ScaleHeterogeneous Mobile Robot Team", Int. Symposium on Experimental Robotics (ISER), 2004.

[HRWI07] G. Huang, A. Rad, Y-K. Wong, Y-L. Ip, "Heterogeneous multisensor fusion for mapping dynamic

environments". Advanced Robotics 21(5-6):661-688, 2007.

[HSCY05] Y.G. Ha, J.C. Sohn, Y.J. Cho, and H. Yoon, "Towards a Ubiquitous Robotic Companion: Design andImplementation of Ubiquitous Robotic Service Framework", ETRI Journal, 27(6):666-676, 2005.

[INPS03] L. Iocchi, D. Nardi, M. Piaggio, and A. Sgorbissa, "Distributed coordination in heterogeneous multi-robotsystems". Autonomous Robots, 15(2):155168, 2003

[KFL+03] K. Konolige, D. Fox, B. Limketkai, J. Ko, and B. Steward, "Map merging for distributed robotnavigation". In Proc. of the IEEE/RSJ Int. Conf. on Intelligent Robots and Systems, pages 212217. 2003.

[KFP07] H. Kress-Gazit, G. Fainekos, and George J. Pappas, "Where's Waldo? Sensor-Based Temporal LogicMotion Planning". IEEE Conference on Robotics and Automation, 2007.

[KH03] A. Khoo, and I. Horswill, "Grounding inference in distributed multi-robot environments". Robotics andAutonomous Systems 43(2-3):121-132, 2003.

[KHEI04] T. Kanda, T. Hirano, D. Eaton, H. Ishiguro, "Interactive Robots as Social Partners and Peer Tutors forChildren: a Field Trial", Human Computer Interaction, 19(1-2):61-84, 2004.

[Kiy03] K. Kiyoshi, "Ubiquitous Intelligent Robotics", ATR UptoDate, no.5, 2003, URL:http://results.atr.jp/uptodate/ATR_2003/sum/kogure.html.

[KKL04] J. H. Kim, Y.D. Kim, and K.H. Lee, "The third generation of robotics: Ubiquitous robot", in Proc. of the2nd Int. Conf. on Autonomous Robots and Agents, 2004.

[KKY+06] Y.G. Kim, H.K. Kim, S.H. Yoon, S.G. Lee, and K. D. Lee, "Home Security Robot based on SensorNetworks", SICE-ICASE Intern. Joint Conf. 2006.

[KSP+06] T. Kanda, M. Shiomi, L. Perrin, T. Nomura, H. Ishiguro, and N. Hagita, "Analysis of People Trajectorieswith Ubiquitous Sensors in a Science Museum", IEEE Int. Conference on Robotics and Automation (ICRA),2007.

[LH02] J. H. Lee and H. Hashimoro, "Intelligent Space Its concept and contents", Advanced Robotics Journal,16(4), 2002.

[LKS07] R. Lundh, L. Karlsson and A. Saffiotti. "Dynamic Self-Configuration of an Ecology of Robots." Proc. ofthe IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS), pp. 3403-3409. San Diego, CA, 2007.

[LOH06] O. Lemaire, K. Ohba, and S. Hirai, "Dynamic integration of ubiquitous robotic systems using ontologiesand the RT middleware", Int. Conf. on Ubiquitous Robots and Ambient Intelligence, 2006.

[LS08] K. LeBlanc and A. Saffiotti. "Cooperative Anchoring in Heterogeneous Multi-Robot Systems." Proc. of theIEEE Int. Conf. on Robotics and Automation (ICRA). Pasadena, CA, 2008.

[LVSM08] P. U. Lima, N. Vlassis, M.T.J. Spaan, and F. S. Melo. Workshop on formal models and methods formulti-robot systems at AAMAS08, May 2008.


40/137

[MC96] M. J. Mataric and D. Cliff, "Challenges in evolving controllers for physical robots". Robotics andAutonomous Systems 19(1):67-83, 1996.

[MCF+07] L. Merino, F. Caballero, J. Ferruz, J. Wiklund, P-E. Forssn, and A. Ollero. Multi-UAV CooperativePerception Techniques, "Multiple Heterogeneous Unmanned Aerial Vehicles", pages 67-110, 2007.

[MVO05] I. Maza, A. Viguria, and A. Ollero, "Aerial and Ground Robots Networked with the Environment", IEEE

Workshop on Network Robot Systems, Int. Conf. on Robotics and Automation (ICRA), 2005.

[Oh03] S. R. Oh, "IT Based Intelligent Service Robot", Proc. First NSF PI Workshop on Robotics and ComputerVision (RCV'03) Invited talk, Las Vegas, 2003, http://www.vcl.uh.edu/~rcv03/materials/slides/SangRok.ppt.

[OSCS07] S.R. Oh, L. Schenato, P. Chen, and S. Sastry, "Tracking and Coordination of Multiple Agents usingSensor Networks: System Design, Algorithms and Experiments", Proceedings of the IEEE, 95(1), 2007.

[OSV08] F. Oliehoek, M. Spaan, and N. Vlassis, "Optimal and approximate Q-value functions for decentralizedPOMDPs". Journal of Artificial Intelligence Research, 2008.

[Par98] L. E. Parker, "ALLIANCE: An architecture for fault tolerant multirobot cooperation". IEEE Transactions onRobotics and Automation, 14(2):220240, April 1998

[Par00] L. E. Parker, "Lifelong Adaptation in Heterogeneous Multi-Robot Teams: Response to Continual Variation

in Individual Robot Performance", Autonomous Robots, 8 (3):239-267, 2000.

[PDM+99] E. Pagello, A. DAngelo, F. Montesello, F. Garelli, and C. Ferrari, "Cooperative behaviors in multi-robotsystems through implicit communication". Robotics and Autonomous Systems, 29(1):6577, 1999

[PFF+03] M. Pilipose, K.P. Fishkin, D. Fox, H. Kautz, D Patterson, and M. Perkowitz, "Guide: TowardsUnderstanding Daily Life via Auto-Identification and Statistical Analysis", Int. Workshop on UbiquitousComputing for Pervasive Healthcare Applications, 2003.

[PM06] J. Pugh and A. Martinoli, "Multi-robot learning with particle swarm optimization". Proceedings of the Fifthinternational Joint Conference on Autonomous Agents and Multiagent Systems, 2006.

[PSK+04] D. Popa, A. Sanderson, R. Komerska, S Mupparapu, D. R. Blidberg, S. Chappel, "Adapting SamplingAlgorithms for Multiple Autonomous Underwater Vehicels", IEEE Autonomous Underwater Vehicles

Workshop, 2004.[SA06] A. Sanfeliu and J. Andrade-Cetto, "Ubiquitous networking robotics in urban settings", Workshop on

Network Robot Systems. Toward Intelligent Robotic Systems Integrated with Environments. IEEE/RSJInternational Conference on Intelligence Robots and Systems (IROS), 2006.

[SA08] A. Sanfeliu, J. Andrade-Cetto, R. Alami, A. Ollero, R. Siegwart, P. Dario, J. Sequeria, J. Santos-Victor, L.Montano, Chapter : Network Robot Systems, RoadMap EURON II (in press), 2008

[Sat02] M. Satyanarayanan, "A C

european robotics research network

Documents