Multi-Robot S y stems. From Swarms to Intelli g ent Automata V olume III Volume III Proceedings from the 2005 International Workshop on Multi-Robot Systems Edited b y LYNNE E. PARKER T h e University o f Tennessee , Knoxvi ll e, TN, U.S.A . an d FRANK E. SCHNEIDER Multi-Robot Systems. From Swarms to Intelligent Automata ALAN C. SCHULTZ N av y Center f or Applied Research in A.I. , N ava l Researc h La b oratory , Was h ington, DC, U.S.A . F GAN, Wac h t b erg, Germany A C.I.P. Catalogue record for this book is available from the Library of Congress. Published by Springer, P.O. Box 17 , 3300 AA Dordrecht , The Netherlands. Printed on acid- f ree pape r All Rights Reserved © 2005 Sprin g e r No part of this work may be reproduced, stored in a retrieval system, or transmitted i n any form or by any means, electronic, mechanical, photocopying, microfilming, r ecording or otherwise, without written permission from the Publisher, with the exception of an y material supplied specificall y for the purpose of bein g entered and executed on a computer system, for exclusive use by the purchaser of the work. Print ed in th e N e th e rlan ds. ISBN-13 978-1-4020-3388-9 (HB) Springer Dordrecht, Berlin, Heidelberg, New York ISBN-10 1-4020-3388-5 (HB) Sprin g er Dordrecht, Berlin, Heidelber g , New York ISBN-10 1-4020-3389-3 (e-book) Springer Dordrecht, Berlin, Heidelberg, New York I SBN-13 978-1-4020-3389-6 (e-book) Springer Dordrecht, Berlin, Heidelberg, New York C ontents Pr e f ace ix Part I Task Allocatio n The Generation of Bidding Rules for Auction-Based Robot Coordination 3 C rai g Tove y , Michail G. La g oudaki s , Sonal Jain, and Sven Koeni g Issues in Multi-Robot Coalition Formatio n 15 L ovekesh Vi g and Julie A. Adam s Sensor Network-Mediated Multi-Robot Task Allocatio n 27 M axim A. Batalin and G aura vS . S ukhatm e Part II Coordination in Dynamic Environment s M ulti-Ob j ective Cooperative Control of D y namical S y stem s 41 Z hihua Q u , Jin g Wan g , a nd Richard A. Hul l L evels of Multi-Robot Coordination for D y namic Environment s 53 C olin P. McMillen, Paul E. R y bski, and Manuela M. Velos o Parallel Stochastic Hill-Climbin g with Small Team s 6 5 B rian P. Gerke y , Sebastian Thru n , Geo ff Gordo n T owar d Versat ili t y o f Mu l t i -Ro b ot S y stem s 79 C o l in C h err y an d Hon g Z h an g Part III In f ormat i on / Sensor S h ar i n g an d Fus i o n Decentra li ze d Commun i cat i on Strate gi es f or Coor di nate d Mu l t i -A g ent Po li c i es 9 3 M aayan Rot h , Rei d Simmons, an d Manue l aVe l os o Improving Multirobot Multitarget Tracking by Communicating Ne g at i ve In f ormat i o n pgg 107 M att h ew Powers, Ramprasa d Ravic h an d ran, Fran k De ll aert, an d Tuc k er Ba l c h vi MU LTI-R O B O T S Y S TEM S Enabling Autonomous Sensor-Sharing for Tightly-Coupled C ooperat i ve Tas k s g 119 Ly nne E. Parker, Maureen Chandra, and Fan g Tan g Part IV Distributed Mapping and Coverag e Merging Partial Maps without Using Odometr y 1 3 3 Distributed Coverage of Unknown/Unstructured Environments b yMo bil e Sensor Networ ks g 14 5 P art V Motion Planning and Contro l 1 59 J ames Bruce and Manuela V elos o A Multi-Robot Testbed for Biologically-Inspired C ooperat i ve Contro l 171 Rafael Fierro, Justin Clark , k k Dean Hou g en, and Sesh Commuri P art VI Human-Robot Interactio n T ask Switching and Multi-Robot Team s 1 8 5 Michael A. G oodric h , Mor g an Qui g le y , a nd Ker y l Cosenz o User Modelling for Principled Sliding Autonomy in Human-Robot Teams 19 7 Brennan Sellner, Reid Simmons, and San j iv Sing h P art VII A pp lication s Multi-Robot Chemical Plume Tracin g 211 Diana Spears, Dimitri Zarzhitsk y , a nd David Tha y er Deploying Air-Ground Multi-Robot Teams i nUr b an Env i ronment s pyg pyg 223 L. Chaimowicz, A. Cowle y , D. Gomez-Ibanez, B. Grocholsk y , M. A. Hsieh , H. Hsu, J . F. Keller, V. Kumar, R. Swaminathan, and C. J. Ta y lo r P art VIII Poster S h ort Paper s A Robust Monte-Carlo Algorithm for Multi-Robot Localizatio n 251 A Dialogue-Based Approach to Multi-Robot Team Contro l 2 5 7 N athanael Chambers, James Allen, Lucian Galescu, and Hyuckchul Jun g F r F F ancesco rr A mi g oni , S imon e G as p arini, and Maria Gin i I oanni s Rekleiti s , Ai P eng PP N ew , and Howie Choset R eal-Time Multi-Robot Motion Planning with Safe Dynamic s Va z ha Amiranashvil i and Gerhard Lakeme y e r Prec i s i on Man ip u l at i on w i t h Coo p erat i ve Ro b ot s 2 35 A s h l e y e S t r o r r u p u e , T e T T r r y r r H u HH n t s tt b e r g rr e r , r r A v i O k o n , and Hrand Aghazarian C ontent s vii for Mobile Robot Teams 263 J ason Derenick, Christo p her Thorne, and John S p letze r T h e G NATs – Lo w - C ost Em b e dd e d Net w or ks f or Support i n g Mo bil eRo b ot s 277 Keit h J. O’Hara, Danie l B. Wa lk er, an d Tuc k er R. Ba l c h 2 9 1 2 9 9 S warm i n g UAVS Be h av i or H i erarc hy 269 K uo- C hi Li n Ro l eBase d Operat i on s 283 B rian Satterˇ eld , Heeten Choxi , and Drew Housten Hybrid f Free-Space Optics/Radio Frequency (FSO/RF) Networks bil b b Er godic Dynamics by Design: A Route to Predictable Multi-Robot System s A ut h or In de x Dy la n A. S hell , C hri s V V V J ones, and Maja J. Matari JJ c ´ Prefac e T h eT hi r d Internat i ona l Wor k s h op on Mu l t i -Ro b ot Systems was h e ld in March 200 5 at the Naval Research Laboratory in Washington, D.C., USA . Br i ng i ng toget h er l ea di ng researc h ers an d government sponsors f or t h ree d ay s of tec h n i ca li nterc h ange on mu l t i -ro b ot systems, t h ewor k s h op f o ll ows tw o p rev i ous hi g hl y success f u l gat h er i ngs i n 2002 an d 2003.L ik et h e prev i ous tw o wor k s h ops, t h e meet i ng b egan w i t h presentat i ons b yvar i ous government pro - gram managers d escr ibi ng app li cat i on areas an d programs w i t h an i nterest in m u l t i -ro b ot systems. U.S. Government representat i ves were on h an df rom t h eO ffi ce o f Nava l Researc h an d severa l ot h er governmenta l o ffi ces.Top re - searc h ers i nt h e fi e ld t h en presente d t h e i r current act i v i t i es i n many areas o f m u l t i -ro b ot s y stems. Presentat i ons spanne d aw id e ran g eo f top i cs, i nc l u d - i n g tas k a ll ocat i on, coor di nat i on i n dy nam i cenv i ronments, i n f ormat i on/senso r s h ar i n g an df us i on, di str ib ute d mapp i n g an d covera g e, mot i on p l ann i n g an d c ontro l , h uman-ro b ot i nteract i on, an d app li cat i ons o f mu l t i -ro b ot s y stems. A ll p resentations were g iven in a sin g le-track workshop format. This proceed - i n g s documents the work presented at the workshop.The research presenta - tions were followed b y panel discussions, in which all participants interacte d to hi g hli g ht the challen g es of this field and to develop possible solutions. I n addition to the invited research talks, researchers and students were g iven a n o pportunit y to present their work at poster sessions.We would like to thank th e Naval Research Laborator y for sponsorin g this workshop and providin g the fa- c ilities for these meetin g s to take place.We are extremel yg rateful to Ma g dalen a Bu g a j ska, Paul Wie g and, and Mitchell A. Potter, for their vital help (and lon g hours) in editin g these proceedin g s and to Michelle Caccivio for providin g th e administrative su pp ort to the worksho p . L YNNE E . P ARKER , A L AN C . S C H U LT Z , A ND F R A N K E . S C HNEIDER ix I T A S K ALL OC ATI ON THE GENERATION OF BIDDING RULES FOR AUCTION-BASED ROBOT COORDINATION ∗ C ra i g Tovey, M i c h a il G. Lagou d a kis Sc h oo l of In d ustria l an d Systems Engineering, Georgia Institute of Tec h no l og y { ctovey, m i c h a il . l a g ou d a kis } @ isye. g atech.ed u S ona l Ja i n, Sven Koen i g Computer Science Department, University of Sout h ern Ca l iforni a { s ona lj a i ,s k oen ig } @ usc.ed u Abs tr act R o b ot i cs researc h ers h ave use d auct i on- b ase d coor di nat i on systems f or ro b o t t eams because of their robustness and efficiency. However, there is no researc h i nto systematic methods for deriving appropriate bidding rules for given tea m o bjectives. In this paper, we propose the first such method and demonstrate it b y d eriving bidding rules for three possible team objectives of a multi-robot explo - r ation task. We demonstrate experimentally that the resulting bidding rules in - d eed exhibit good performance for their respective team objectives and compar e f avorably to the optimal performance. Our research thus allows the designer s o f auction-based coordination systems to focus on developing appropriate tea m o bjectives, for which good bidding rules can then be derived automatically . K eywords: A uctions, Bidding Rules, Multi-Robot Coordination, Exploration . 1. Introduction T h et i me requ i re d to reac h ot h er p l anets ma k es p l anetary sur f ace exp l orat i o n mi ss i ons pr i me targets f or automat i on. Sen di ng rovers to ot h er p l anets e i t h e r i nstea d o f or toget h er w i t h peop l e can a l so s i gn ifi cant l yre d uce t h e d anger an d c ost i nvo l ve d . Teams o f rovers are b ot h more f au l tto l erant (t h roug h re d un - d ancy) an d more e ffi c i ent (t h roug h para ll e li sm) t h an s i ng l e rovers if t h e rover s are coor di nate d we ll . However, rovers cannot b e eas il yte l e-operate d s i nce t his ∗ W et h an k Apurva Mu dg a lf or hi s h e l p. T hi s researc h was part ly supporte dby NSF awar d sun d er contract s I TR/AP0113881 , IIS-0098807 , and IIS-0350584. The views and conclusions contained in this document are t h ose o f t h e aut h ors an d s h ou ld not b e i nterprete d as represent i n g t h eo ffi c i a l po li c i es, e i t h er expresse d or i mp li e d ,o f t h e sponsor i n g or g an i zat i ons, a g enc i es, compan i es or t h e U.S. g overnment . 3 L.E. Parker et al. (eds.) , M ulti-Robot Systems. From Swarms to Intelligent Automata. Volume III , 3 – 14 . c 2005 Springer. Printed in the Netherlands . 4 T ove y , et al. r equ i res a l arge num b er o fh uman operators an di s commun i cat i on i ntens i ve , e rror prone, an d s l ow. Ne i t h er can t h ey b e f u ll y preprogramme d s i nce t h e ir a ct i v i t i es d epen d on t h e i r di scover i es. T h us, one nee d stoen d ow t h em w i t h t he c apa bili ty to coor di nate autonomous l yw i t h eac h ot h er. Cons id er, f or exam - pl e, a mu l t i -ro b ot exp l orat i on tas k w h ere a team o fl unar rovers h as to v i s i t a num b er o f g i ven target l ocat i ons to co ll ect roc k samp l es. Eac h target must be vi s i te db yat l east one rover. T h e rovers fi rst a ll ocate t h e targets to t h emse l ves, a n d eac h rover t h en v i s i ts t h e targets t h at are a ll ocate d to i t. T h e rovers k no w t h e i r current l ocat i on at a ll t i mes b ut m i g h t i n i t i a ll y not k now w h ere o b stac l e s a re i nt h e terra i n. It can t h ere f ore b e b ene fi c i a lf or t h e rovers to re-a ll ocate t he targets to t h emse l vesast h ey di scover more a b out t h e terra i n d ur i ng execut i on , f or examp l e, w h enarover di scovers t h at i t i s separate dby a big crater f ro m i ts next tar g et. S i m il ar mu l t i -ro b ot exp l orat i on tas k sar i se f or m i ne sweep i n g, searc h an d rescue operat i ons, po li ce operat i ons, an dh azar d ous mater i a l c l ean - i n g , amon g ot h ers . Multi-robot coordination tasks are t y picall y solved with heuristic method s since optimizin g the performance is often computationall y intractable. The y a re often solved with decentralized methods since centralized methods lack ro - bustness: if the central controller fails, so does the entire robot team. Marke t mechanisms, such as auctions, are p o p ular decentralized and heuristic multi - r obot coordination methods (Rabideau et al., 2000). In this case, the robot s a re the bidders and the tar g ets are the g oods up for auction. Ever y robot bid s o n tar g ets and then visits all tar g ets that it wins. As the robots discover mor e a bout the terrain durin g execution, the y run additional auctions to chan g eth e a llocation of tar g ets to themselves. The resultin g auction-based coordinatio n s y stem is efficient in terms of communication (robots communicate onl y nu - meric bids) and com p utation (robots com p ute their bids in p arallel). It is there - f ore not surprisin g that auctions have been shown to be effective multi-robot c oordination methods (Gerkey and Matar i ´ c, 2002, Zlot et al., 2002, Thayer ´ e t al., 2000, Goldberg et al., 2003). However, there are currently no systemati c methods for deriving appropriate bidding rules for given team objectives. I n t hi s paper, we propose t h e fi rst suc h met h o d an dd emonstrate i t b y d er i v i n g biddi ng ru l es f or t h ree poss ibl e team o bj ect i ves o f t h emu l t i -ro b ot exp l orat i o n tas k .We d emonstrate exper i menta ll yt h at t h e resu l t i ng biddi ng ru l es i n d ee d e x hibi t goo d per f ormance f or t h e i r respect i ve team o bj ect i ves an d compare f a - v ora bl ytot h e opt i ma l per f ormance. Our researc h t h us a ll ows t h e d es i gners o f a uct i on- b ase d coor di nat i on systems to f ocus on d eve l op i ng appropr i ate tea m obj ect i ves, f or w hi c h goo d biddi ng ru l es can t h en b e d er i ve d automat i ca ll y . [...]... schedule is not available in advance, and 3 robots need to physically visit task locations to accomplish task completion (e.g., to push an object) Our approach to OMRTA relies on a computational and sensing fabric of networked sensors embedded into the 27 L.E Parker et al (eds.), Multi-Robot Systems From Swarms to Intelligent Automata Volume III, 27–38 c 2005 Springer Printed in the Netherlands 28... et al (eds.), Multi-Robot Systems From Swarms to Intelligent Automata Volume III, 15–26 c 2005 Springer Printed in the Netherlands 16 Vig and Adams coalitions Choosing the optimal coalition from all possible coalitions is an intractable problem due to the size of coalition structure space (Sandholm et al., 1999) Algorithms exist that yield solutions within a bound from the optimal and are tractable... environments onto eight-connected uniform grids of size 51 × 51 and computed all costs between locations as the shortest distances on the grid Our auction-based coordination system used these costs to find an 10 Tovey, et al allocation of targets to robots and a path for each robot that visits all targets allocated to it We interfaced it to the popular Player/Stage robot simulator (Gerkey et al., 2003) to execute... applicability to the multiple-robot domain This work aims to correct that discrepancy by unearthing issues that arise while attempting to tailor these algorithms to the multiple-robot domain A well-known multipleagent coalition formation algorithm has been studied in order to identify the necessary modifications to facilitate its application to the multiple-robot domain Keywords: Coalition formation, fault-tolerance,... fault-tolerance, multi-robot, task allocation 1 Introduction Multi-agent systems often encounter situations that require agents to cooperate and perform a task In such situations it is often beneficial to assign a group of agents to a task, such as when a single agent cannot perform the tasks This paper investigates allocating tasks to disjoint robot teams, referred to as 15 L.E Parker et al (eds.), Multi-Robot. .. performance between known and unknown environments is at most a factor of three It is remarkable that our auction-based coordination system manages to achieve such a good performance for all team objectives since there has to be some performance degradation given that we switched both from known to unknown environments and from eight-connected to fourconnected grids 6 Conclusions and Future Work In this... tasks may be located at considerable distances from one another so that the best solution is to dispatch a robot team to each designated task area and hope that the team can autonomously complete the task The robots must then coalesce into teams responsible for each task The focus of this work is to investigate the various issues that arise while attempting to form multiple-robot coalitions using existing... objective of the multi-robot exploration task is to find an allocation of targets to robots and a path for each robot that visits all targets allocated to it so that the team objective is achieved Note that the robots are not required to return to their initial locations In this paper, we study three team objectives: M INI S UM: Minimize the sum of the path costs over all robots 6 Tovey, et al M INI... experimentally that the derived bidding rules indeed exhibit good performance for their respective team objectives and compare favorably to the optimal performance In the future, we intend to adapt our methodology to other multi-robot coordination tasks For example, we intend to study multi-robot coordination with auction-based coordination systems in the presence of additional constraints, such as compatibility... communicate with each other and are aware of all tasks to be performed Each agent has a vector of real non-negative capabilities Bi =< bi , bi , bi > Each capability quantifies the ability to perform an r 1 2 action In order to assess coalitions and task execution, an evaluation function is attached to each capability type that transforms capability units into monet tary units It is assumed that there is a . Multi-Robot S y stems. From Swarms to Intelli g ent Automata V olume III Volume III Proceedings from the 2005 International Workshop on Multi-Robot Systems Edited b y LYNNE. E. SCHNEIDER Multi-Robot Systems. From Swarms to Intelligent Automata ALAN C. SCHULTZ N av y Center f or Applied Research in A.I. , N ava l Researc h La b oratory , Was h ington, DC, U.S.A . F GAN,. (eds.) , M ulti-Robot Systems. From Swarms to Intelligent Automata. Volume III , 3 – 14 . c 2005 Springer. Printed in the Netherlands . 4 T ove y , et al. r equ i res a l arge num b er o fh uman operators an di s