RFID Digital Pheromones for Generating Stigmergic Behaviour to Autonomous Mobile Robots

Transcription

1 RFID Dgtal Pheromones for Generatng Stgmergc Behavour to Autonomous Moble Robots I. SUSNEA *, G. VASILIU *, A. FILIPESCU * * Department of Automaton and Industral Informatcs Unversty Dunarea de Jos of Galat Str. Domneasca, 47, Galat, , ROMANIA Abstract: Durng the past years, the prce of RFID tags have constantly dropped, whle ther techncal performances ncreased. Along wth the constant nterest of the researchers for the problems of swarm ntellgence, ths opens a wde area of applcatons based on the use of RFID tags as storage meda for dgtal pheromones, amed to medate the communcaton between varous agents. A smple method to defne dgtal pheromones as data structures and to smulate the evaporaton process of the natural pheromones s descrbed. Ths paper descrbes an experment where RFID tags storng dgtal pheromones, deployed n an ndoor envronment, are used to defne a dynamc path to be followed by autonomous moble robots. The robots sense the pheoromones n the envronment by means of a dual antenna, just lke ants do, and a fuzzy controller governs the left-rght movement thereof. The expermental results prove that the dea of usng dgtal pheromones to control the movement of autonomous moble robots s entrely feasble, even wth very smple robots. Ths opens new possbltes to mprove robot to robot and human to robot nteracton. The dea of enablng moble robots wth new senses, and thus gong beyond the antropomorphc paradgm n robot control, seems to be productve. Keywords: Dgtal pheromones, RFID, moble robots, fuzzy controller, stgmergy.. Introducton The performances of the RFID technology are so mpressve, that some researchers, lke Franco Zambonell et al. n [] envson the possblty of sprayng mcrocomputers n the envronment to create complex, self-organzng structures. Ths llustrates the growng nterest of the researchers for the swarm ntellgence and stgmergy. A comprehensve study [2], conducted by Tony Whte for the Canadan Department of Natonal Defence, dentfes hundreds of sgnfcant works n ths feld, and there are many others. Whle the general topcs of swarm ntellgence and RFID are very well covered n the lterature (see for example [3] and [4]), there are however relatvely few studes drectly descrbng the use of RFID tags as a means to store dgtal pheromones amed to gude moble robots for tasks lke path plannng, object trackng, or other cooperatve tasks. One of the most nterestng works wth ths subject s [5], whch deals wth the problem of object trackng by autonomous moble robots. Ths paper s partcularly nterestng because t offers an elegant soluton to the problem of dgtal pheromone renforcement/evaporaton, by addng a tme stamp to the data structure whch defnes the ntensty of the pheromone. Ths tme stamp s read by the agent that senses the pheromone, and compared wth ts own real tme clock. Based on the age of the latest record, each agent computes and updates the ntensty of the pheromone. Ths paper descrbes an experment amed to prove that RFID tags deployed n the envronment, storng dgtal pheromones, can easly defne paths to be followed by autonomous robotc agents. Robots are controlled n a way that emulates the foragng behavour of the socal nsects. Ths smple soluton can be, n some crcumstances, a valuable alternatve to complex envronment mappng and path plannng algorthms. Our experment s, at ths tme, lmted to a software smulaton, but ths smulaton uses the knematc model of the real robot Poneer3-DX, from MobleRobots Inc. [6], and the whole applcaton s wrtten n ANSI C, so that t s easly portable to an embedded controller carred by the robot. In secton 3, we nclude the schematc of a possble embedded structure canable to execute the tasks descrbed here. The paper s structured as follows: Secton 2 ntroduces the concept of dgtal pheromone and shows the RFID mplementaton thereof. Secton 3 presents the knematc model of the robot and suggests ISSN: ISBN:

2 a possble structure of an embedded devce sutable for a phscal mplementaton of the experment. In secton 4 s presented the fuzzy controller that governs the movement of the robot. Secton 5 presents the expermental results, whle secton 6 s reserved for general dscussons and conclusons. 2. RFID tags and dgtal pheromones Bascally, RFID tags contan a nonvolatle memory (usually 0.5-8k EEPROM), whose content s accessble to a so-called reader va a contactless rado lnk. Both read and wrte operatons are possble. Ther man advantages are that they are powered by the rado frequency wave generated by the reader, so that they do not have the battery-exhauston problem, and the very low cost. A reader can access the nformaton stored n a RFID tag from a dstance rangng from a few centmeters (ISO4443) to up to m (ISO5693). See detals n [4] and [7]. At a cost of only a few cents per tag, hundreds or thousands of such tags can be deployed n the envronment, and a varety of agents human or robotc carryng approprate readers can access the nformaton stored n the nternal memory of the tags. Pheromones are chemcal markers that most socal nsects (and other speces) spread n the envromnemt n order to send messages and modulate the behavour of the other ndvduals of the colony. The term pheromone was ntroduced n 959 by Karlson and Luescher n [8], and the mechansms of nter-ndvdual coordnaton between ants by means of pheromones was explaned n 959 by Perre-Paul Grassé n [9]. Ant foragng s the most common example of pheromone-medated ntercoordnaton. When an ant fnds food, t starts spreadng pheromone on ts way back to the nest, thus creatng a tral that ndcates the way to the food source to the other ants. Each ant usng the same path renforces the pheromene tral. When the food source s exhausted, n the absence of renforcement, the natural evaporaton makes the tral dsappear. Ths process of ndrect coordnaton between agents by means of a trace left n the envronment by an acton, whch stmulates smlar subsequent actons s called stgmergy (Grassé [9]). RFID tags seem to be the deal soluton to create dgtal pheromones for use wth robotc agents and other applcatons of stgmergy. However, due to ther passve nature, RFID tags cannot smulate by themselves the process of natural evaporaton. Therefore, we have adpoted the followng soluton: the actual dgtal pheromone conssts n a counter, whch can be ncremented or decremented, and a tme stamp. Every tme an agent accesses the nformaton n the tag, t reads ths tme stamp, compares t wth ts own real tme clock and decdes whether to ncrement or decrement the counter, based on the age of the current record. Then, t overwrtes the record, storng the computed value of the counter, and a new tme stamp. Fgure, shows the data structure stored by the RFID tags, llustratng the dea of the mplementaton of dgtal pheromones. AGENT ID TYPE INTENSITY COUNTER TIME STAMP Fg. Data structure defnng the dgtal pheromone The counter that gves the ntensty of the dgtal pheromone s read and updated by every agent at dscrete tme moments T, accordng to the expresson (). Cn+ = Cn + Qf ( Tn+ Tn ) () where: for t< t + f ( t) = T + T (2) n n nt( ) for t> t t Q s a postve constant, representng the quantum of renforcement/evaporaton of the pheromene nt(x) s the nteger part of x, and t s the evaporaton constant. T n+ s the current tme accordng to the real tme clock of the agent, and T n s the tme stamp of the prevous record, read from the tag. The feld Agent ID s was ncluded to prevent multple successve renforcements of the pheromone by a slow movng agent, and the feld Type s reserved for future use, because t may be nterestng n some stuatons to defne multple types of pheromones to nduce dfferent behavours to the populaton of agents. 3. The robotc agents Real ants sense the pheromone tral by means of two antennas, and they constantly correct ther trajectory accordng to the dfferental ntensty of the pheromene smell. We propose a smlar sensng structure for the robotc agents, wth two laterally placed RFID readers, as shown n fgure 2. Wth ths placement of the RFID readers, each robot wll create/renforce a double tral of dgtal pheromones. Ths mproves the overall robustness of ISSN: ISBN:

3 the system to the stuatons of broken pheromone trals. Y Reader Left b V L (x,y) V V R Reader Rght Fg. 2 Knematc schematc of the robot Our smulaton program uses the knematc model of the Poneer3 robot presented n detal n our prevous work [0]. Wth the notatons n fgure 2, ths s snthetcally expressed n (3). snφ snφ x' 2 2 cosφ cosφ vl = y' (3) 2 2 vr φ ' b b The actual control of the Poneer3 robot s performed by sendng commands va the seral communcaton nterface, accordng to a propretary protocol. The command VEL2, for example, allows the control unt to specfy dstnct values for the speed of the left and rght wheels of the robot. Wth ths feature, a command VEL2(v L >v R ) means rght turn, VEL2(v L <v R ) means left turn, and VEL2(v L =v R ) means smple translaton forwards. A possble structure for the embedded control unt, sutable for ths applcaton s presented n fgure 3. MCU FUZZY CONTROL MCU2 SENSING RFID READER X The suggested structure of the embedded controller of the robot uses two mcrocontrollers: MCU controls the moton of the robot by means of a fuzzy controller and sends commands to the robot va the RS232 nterface, whle MCU2 holds the real tme clock of the system and perodcally sends nterrogatons to the RFID readers. The computed values of the ntensty of the pheromone are presented to the man mcrocontroller as two analog sgnals. 4. The Fuzzy Controller A smple fuzzy controller can solve the problem of the robot movement by processng a set of logc statements lke ths: f the pheromone level sensed by the rght reader s, and the level of pheromone sensed by the left reader s, then v L must be and v R must be. Assumng that there are only two fuzzy subsets ( and ) for the doman of varaton of the pheromone level, one can choose membershp functons lke those depcted n fgure 4. DOM 00% LT INTENSITY Fg. 4 Membershp functons for pheromone levels Wth ths assumpton, the whole rule base that descrbes the behavour of the fuzzy controller s summarzed n table. Table The rule base of the fuzzy controller LEFT LEVEL HT RIGHT LEVEL LL LH HL HH RFID READER 2 Fg. 3 Suggested structure of the embedded controller ISSN: ISBN:

4 The fuzzy controller generates crsp values for v L and v R by processng the fuzzy rule base, accordng to (4). v K = OUT = K = z S z (4) where: z = mn( PIL, PIR ) (5) S s the correspondng sngleton value of the fuzzy output, and K s the total number of rules n the rule base. PIL, PIR are the degrees of membershp of PIL(t) and PIR(t) to the doman correspondng to the cell. PIL(t) and PIR(t) are the pheromone ntensty functons for the left and rght reader respectvely. (see [0] and [] for detals on the mplementaton of the fuzzy controller). 5. Expermental Results General assumptons a. The RFID tags are unformly dtrbuted n the envronment, at a dstance so that one and only one tag can be accessed at any moment by each of the readers. b. An ntal pheromone tral s marked on the pheromone array, servng as target path for the robots (see fgure 5) Fgure 6 depcts the trajectory of the frst robot crossng the pheromone area, by showng the successve postons of the axs connectng the two RFID readers as dark segments. START Fg. 6 Trajectory of the frst robot After havng 50 robots crossng the pheromone area, and consderng the effect of evaporaton, the pheromone tral looks lke the mage n fgure 7. Fg. 7 Pheromone tral after 50 robots Fg. 5 Intal condtons of the experment Note that n fgure 5 and subsequent each cell of the array contans a RFID tag. Whte cell have the concentraton of pheromone equal to zero. Gray cell have hgher pheromone concentratons the darker shades ndcate hgher concentratons. And the trajectory of a robot through ths envronment s much closer to a straght lne (see fgure 8). Note that better performances can be easly obtaned by mprovng the fuzzy controller, e.g. by choosng three fuzzy subsets (Low, Medum, Hgh) for the doman of varaton of the concentraton of pheromone. Also, n some applcatons, t s possble to abandon the strct stgmergy scenaro and to have a human operator defne more complex ntal condtons ISSN: ISBN:

5 for the experment by creatng thcker pheromone trals. START Fg. 8 Trajectory wth hgher pheromone densty 6. Dscusson Ths smple experment clearly demonstrates that RFID based dgtal pheromones can be easly used to defne dynamc paths for gudng moble robots. In some stuatons, ths can be an effcent and cost effectve alternatve to complex envronment mappng and path plannng algorthms. It s part of our research effort amed to develop low cost embedded solutons for robot control. The man weak pont of ths experment s that t s lmted to a software smulaton, but we beleve t can be easly reproduced wth embedded hardware. Further research s requred to dversfy the robotenvronment nteracton wth new scenaros (e.g. danger ahead don t cross ths lne ) or to mprove human - robot nteracton. Snce the memory capacty of the commercally avalable RFID tags s now much hgher than the few bytes requred by the dgtal pheromone used n ths experment, t would be possble to use ths resource ether by defnng dfferent types of pheromones, or by storng more complex mappng nformaton about the envronment. Ths experment only smulates the foragng behavour of the ants n a robotc context, but there are much more aspects concernng swarm ntellgence that can be explored wth ths smple expermental setup. References [] Zambonell Franco, Glezes Mare-Perre, Mame Marco, Tolksdorf Robert - Spray Computers: Exploratons n Self-Organzaton, n Journal of Pervasve and Moble Computng, Vol., No., pp [2] Whte Tony Expert Assessment of Stgmergy, A Report for the Department of Natonal Defence, Contract No. W /003/SV, Fle No. 0 sv.w , Requston No. W [3] Bonabeau E., Dorgo M., Theraulaz G., Swarm Intellgence, Oxford Unversty Press (Oxford, UK), 999 [4] Fnkenzeller Klaus, RFID Handbook John Wley and Sons 2003, ISBN [5] Mame, Marco. Zambonell, Franco - Physcal deployment of dgtal pheromones through RFID technology, Swarm Intellgence Symposum, SIS Proceedngs 2005 IEEE [6] Moble Robots Inc. [7] Rankl Wolfgang, Effng Wolfgang - Smart Card Handbook, John Wley and Sons 2003, ISBN [8] Karlson, P., Lüscher, M.. Pheromones: a new term for a class of bologcally actve substances. Nature 83 (959), pp [9] Grassé Perre-Paul. La Reconstructon du nd et les Coordnatons Inter-Indvduelles chez Bellcostermes Natalenss et Cubtermes sp. La theore de la Stgmerge: Essa d nterpretaton du Comportement des Termtes Constructeurs. Insectes Socaux, 6:48, 959 [0] Susnea Ioan, Vaslu Grgore, Flpescu Adran - Real-tme, embedded fuzzy control of the Poneer3- DX robot for path followng, Proceedngs of the WSEAS Conference n Crete, 2008 [] Susnea I. and Mtescu M. Mcrocontrollers n Practce, SPRINGER Verlag, 2005, ISBN ISSN: ISBN: