Localization Algorithm for a mobile robot using igs

Transcription

1 Poceedings of the 17th Wold Congess The Intenational Fedeation of Automatic Contol Seoul, Koea, July 6-11, 008 Localization Algoithm fo a mobile obot using igs DaeGeun Seo *, JangMyung Lee** * Electical Engineeing Depatment, Pusan National Univesity, Pusan, Koea, ( sdk@pusan.ac.k). ** Electical Engineeing Depatment, Pusan National Univesity, Pusan, Koea, ( jmlee@pusan.ac.k)} Abstact: As an absolute positioning system, igs is designed based on ultasonic signals whose speed can be fomulated clealy in tems of time and oom tempeatue, which is utilized fo mobile obot localization. The igs is composed of an RFID eceive and an ulta-sonic tansmitte, whee an RFID is designated to synchonize the tansmitte and eceive of the ultasonic signal. The taveling time of the ultasonic signal has been used to calculate the distance between the igs system and a beacon which is located at a pe-specified location. This pape suggests an effective opeation method of igs to estimate position of the mobile obot woking in an unstuctued envionment. To expand ecognition ange and to impove accuacy of the system, two stategies ae poposed: auto-calibation of beacons belonging to neighboing blocks and emoval of the envionment-eflected ultasonic signals. As the esults, the ubiquitous localization system based on the igs as a pseudo-satellite system has been developed successfully with a low cost, a high update ate, and elatively high pecision. 1. INTRODUCTION Recently ubiquitous envionment has been applicable fo the home and factoy automation. As one of the typical applications, autonomous navigation system is being developed with vaious sensos fo ecognizing the suounding envionment. Fo the autonomous navigation, the position of the navigation system must be ecognized pecisely and quickly (S.M.LEE et.al., 001). Thee ae seveal indoo position ecognition systems such as IR senso based system, ultasonic system, and camea based system. Among the vaious sensos system, ultasonic senso system is widely used since the system is cheap and simple to implement while it is neithe applicable fo long distance measuing no obust against the eflected waves (H.Pemans et.al., 1993). Thee ae seveal location ecognition systems using ultasonic sensos such as Active Bat, igs, etc. Thee have been seveal eseaches on this eseach whee thee o moe beacons in a small oom ae utilized fo the location ecognition (A.M Flynn, 1998 and Tatsuo Aail et.al., 1983). Howeve cuently fo public sevices, the mobiles ae equied to move in a wide space. The task is not feasible using the ultasonic sensos which ae sensitive to noises and have a shot effective distance. Theefoe in this pape to apply igs not in a small oom but in a wide sevice aea, a block expansion algoithm is intoduced. When the mobile obot is moving to a new block, a new set of beacons needs to be selected to get the position data moe pecisely instead of keeping the same set of beacons. Selecting the new set of beacons, eliminating eflected signal effects against obstacles, and educing the envionment noise effects ae majo poblems to be solved fo the obust contol of mobile obot in a wide aea. Section intoduces basic pinciples of indoo location ecognition using ultasonic sensos and igs which ae used fo eal expeiments. Section 3 poposes a new auto calibation algoithm to ecognize the beacon coodinates when the mobile obot is moving into a new block, which educes the localization time and impoves the data eliability. Section poposes NRW algoithm to esolve the sensitiveness of ultasonic sensos to envionment noise and eflected wave. Section 5 veifies effectiveness of the poposed algoithms though the eal expeiments..1 Distance measuement. Location Measuement Basically the distance measuement can be done by measuing taveling time of the ultasonic signal and its velocity. The ultasonic wave TOF (taveling Time Of Flight) can be usually detected by using the theshold value as shown in (Fig. 1) fo the eceived signal. The tansmission equest signal is sent out fom the tansmitte at T t, and the ultasonic signal is eceived at T. The diffeence between the two epesents the taveling time of the ultasonic signal fom a beacon to the mobile obot (S.Hong et.al., 00). TOF = T T (1) t Fig. 1. Time Of Flight /08/$ IFAC / KR

2 17th IFAC Wold Congess (IFAC'08) Seoul, Koea, July 6-11, 008 The velocity of the ultasonic wave, c, is dependent on the absolute tempeatue, T, and it is epesented as T c [ m / s ] () 37 Even though the velocity may depend on othe factos of the tansmission channel, (), is a geneal fom fo ultasonic wave velocity (Billu Bashan, 000 and M.Paialla et.al., 1991). In this eseach, the woking envionment of the mobile obot is kept at oom tempeatue of nomally. Theefoe the velocity, c, is eplaced by 33m/s to speed up the measuement pocess. And the distance, d, fom the mobile obot to a specific beacon can be calculated as the multiplication of the velocity and the taveling time as. igs(indoo GPS system) d = c TOF (3) As an indoo localization system, igs has been used in this eseach, which is developed by PNU Intelligent Robot Lab. and Ninety Systems Cop. As illustated in (Fig. ), a localize which is installed on the mobile obot sends out an RF ID signal to call a beacon. When a coesponding beacon eceives the RFID signal, it sends out the ultasonic wave signal to the localize. The TOF can be detemined by counting the time fom the moment of RFID tansmission to the moment of ultasonic wave aival at the localize. Since the localize is located at a point in thee dimensional space, if thee ae thee distant data fom diffeent beacons at a moment, the thee dimensional position coodinates can be calculated by the tiangulation technique. Fom the geneal tiangulation technique illustated in (Fig. 3), the absolute coodinates of the mobile obot the localize-- can be detemined uniquely. beacon. EKF (Extended Kalman Filte) is utilized to impove the localization accuacy (S.B. Kim et.al., 007). Fig.. Opeating sequence of the igs and localize. When the localize sends out a RFID signal, the counte stats to count, and it stops when it eceives the ultasonic signal fom the coesponding beacon. Ignoing the taveling time of RFID, the taveling time of the ultasonic signal can be measued by the counte. Using the counte value and the taveling speed which is deteministic, the distance fom the localize to the coesponding beacon can be calculated as explained peviously. Based on the distance values to the beacons which ae holding thei own position data, the location of the mobile obot can be calculated. When the distances fom the localize to the beacons ae epesented as d 1, d and d 3, and the absolute positions of the beacons ae ( x1, y1, z 1), ( x, y, z) and ( x 3, y3, z3 ), the location coodinates of the mobile obot, ( x, y, z ), can be obtained as an intesection of thee cicles illustated in (Fig. 3). Fom the cicle equations, the mobile obot coodinates can be calculated as follows: ( x x1) ( y y1) ( z z1) d 1 ( x x) ( y y) ( z z) = d ( x x3) ( y y3) ( z z3) d 3 () Fo the beacon 1, (), can be expanded. d 1 x y z x x1 y y1 z z1 x1 y1 z1 =. () Fig.. Composition of obot localization system. This fom of thee equations fo beacon 1, and 3 can be eodeed fo unknown mobile obot coodinates, ( x, y, z) x y z x x y z x y z d x y z y = x y z x y z d x y z z x y z x y z d Now the unknown mobile obot coodinates can be calculated as 1 X = A B (7) (6) Fig. 3. Geneal tiangulation technique. (Fig. ) illustates communication flow between igs and the localize of igs. To extend battey lifetime of beacons, the sleep mode is utilized until the coesponding RFID comes into the whee x x y z X y A x y z =, = z x3 y3 z3 73

3 17th IFAC Wold Congess (IFAC'08) Seoul, Koea, July 6-11, 008 and x y z x y z d B = x y z x y z d x y z x y z d Multi Block Scheme with Auto Calibation 3.1 Auto Calibation Algoithm (8) x x y z X = y, A = x y z and z x y z x y z x y z d B = x y z x y z d x y z x y z d (10) Fo the absolute localization pocess, the beacons ae installed on the pe-detemined position of ceiling o walls. Howeve it is vey difficult to attach the beacon pecisely at a specific position since installing beacons on the ceiling and walls is not an easy task. To emove out this had and eoneous installation pocess of the beacons and to obtain eliable absolute position data, an auto calibation algoithm has been developed (S.B.Kim et.al, 006a). With this auto calibation algoithm, the location of the mobile obot can be estimated quickly and pecisely when it is moving out and into a new block as illustated in (Fig. 5) (S.B.Kim et.al, 006b). While the mobile obot is moving in the block, the position of the mobile obot is updated by the thee beacon data. When it moves into block 3, the coodinates of beacon 5 is obtained by the same auto calibation pocess. By using this block expansion and auto calibation algoithms, a mobile obot can move aound wide aea with eliable and speedy position ecognition. 3. Block expansion Anothe shotage of using the ultasonic senso fo the indoo localization is that the ultasonic signal attenuates quickly and becomes uneliable when it is taveling in the noisy envionment. Since the maximum taveling distance fo the ultasonic signal of igs is 15 M, the block expansion technique using multiple beacons is essential fo public sevice aeas such as aipot and tain stations. Fo these applications, as an examples the wide aea is divided into 9 blocks o cells in mobile communication system in (Fig. 6). The block numbe is necessay to be minimized fo cost effectiveness while it is lage enough to povide accuacy and eliability. Fig. 5. Block expanding using auto calibation Thee ae thee blocks each of which is composed of thee beacons fo the block expanding expeiments. Evey 5,000 mm, a beacon is installed at the top of a poll in (Fig. 5). At the initial stage of the expeiment, the mobile obot goes though the auto calibation pocess by locating the mobile obot on the thee x maked positions whose absolute coodinates ae pecisely specified. With the pocess, the mobile obot has its own position coodinates. When the mobile obot is moving into the block, coodinates of the new beacon 3 can be calculated though the auto calibation pocess using the cuent location infomation of the mobile obot and the absolute position infomation of beacons and. This fom of thee equations fo beacon, and 3 can be eodeed fo unknown mobile obot coodinates, ( x, y, z). x y z x x y z x y z d x y z y = x y z x y z d x y z z x y z x y z d Now the unknown mobile obot coodinates can be calculated as (9) Fig. 6. Stuctue of block expansion Fo the given multiple blocks, an ID needs to be assigned uniquely to each individual block, which is illustated in (Fig. 7). Fig. 7. ID fomat of beacons The beacon ID s in the same block ae defined to have close codes to distinguish fom the beacon ID in a new block easily. The oveall flowchat of block expansion algoithm is shown in (Fig. 8) whee a new beacon is coming to be used fo the localization when the mobile obot is moving into a new block. 7

4 17th IFAC Wold Congess (IFAC'08) Seoul, Koea, July 6-11, 008 the diect signals ae hided behind the elatively lage noise signals. The noise signals can be geneated by the eflection of the pevious signal to some obstacles located in the signal path. O it can be geneated by othe envionmental souces like music and constuction sound. The mobile obot needs to be obust against suounding obstacles and envionment, which is the final goal of NRW algoithm. In the algoithm, the pevious flight distance data of the ultasonic signal have been utilized fo estimating next flight distance. Fig. 8. Flowchat of block expansion algoithm The next block fo the mobile obot is estimated fom the velocity vecto of the mobile obot, which shotens the auto calibation time. Fo the optimal auto calibation, thee beacons ae assigned to each block. If thee exists obstacles which pevent the sonic signal taveling fom a beacon to the localize while the mobile obot is moving in a block identification, the Flat Floo Algoithm(S.B.Kim et.al, 006b) fo missing one beacon data and Linea Incemental Algoithm(S.B.Kim et.al, 006a) fo missing two beacon data have been applied fo the tempoay localization. Fo the maximum sixteen blocks along x-axis and y-axis, espectively, a fou bits data fame has been assigned. Theefoe the mobile obot can move aound the maximum 56 blocks. When a block size is 5m 5m, the mobile obot can wok in 80m 80m space ecognizing its own position pecisely.. Random Noise Filteing.1 Eo fom Envionment Noise Fig. 10. Windows data of i th beacon The diffeence between two neighboing flight distance is small enough to be within the toleance eo of igs, 5 Cm, when the mobile obot is moving not so fast, which povides the flight distance ange of ultasonic signal fo each step. The flight distance data in the pe-specified window can be utilized to estimate the next step flight distance, since the mobile obot motion is not abuptly changing. (Fig 10) illustates the data fomat of flight distance stoed in the window, whee d Bij is eceived n th flight distance data, i is the beacon identification numbe, j is the sequence numbe of beacon which is limited to as twenty consideing the data pocessing speed. When d ij is measued, the value is checked whethe that is eliable o not by NRW algoithm as toleance > d i( j 1) dij, d ij= dij othewise, d ij = d ij (11) Fig. 9. Effect of noise The ultasonic signal geneated fo the pevious RFID can be detected as a signal befoe the cuent signal is coming. O the noise signals may aive at the eceive as a signal as illustated in (Fig. 9). To emove out these malfunctioning situations, the gap between two RFID signals can be extended and the theshold value can be inceased not to eceive the small noises as signals. Howeve these teatises limit the woking space and deteioate the data acquisition eliability by lengthening the data pocessing time.. NRW Algoithm An NRW (Noise Rejection Window) algoithm is poposed in this pape to ovecome the malfunctioning situations whee whee d ij is the esult of NRW algoithm. The estimated flight distance of beacon i at the moment of j, d $ ij can be detemined as d d σ = (1) ij i( j 1) i 1 1 n σ i = x m w j= n w ij (13) whee σ is standad deviation of the data stoed in the i window, w is the window size, xij is the vaiance of the data in the window, m is the aveage of the data stoed in the window, d i( j n) is the n step pevious flight distance data of the i th beacon.to obtain the estimated value, standad deviation of the flight distance data stoed in the window fo each beacon has been obtained. As shown in (13),, the value is added to the pevious window data fo the beacon to estimate d ˆij. 75

5 17th IFAC Wold Congess (IFAC'08) Seoul, Koea, July 6-11, 008 envionment in (Fig. 1), the active beacons ae installed fo location ecognition and one additional beacon is placed to geneate noisy signals andomly. Fo the auto calibation pocess, the mobile obot is placed at the pe-detemined locations, A(100,101), B(750,01), and C(500,301) in a sequence. With the pocess, the coodinates of beacons ae obtained as beacon 1 (-15,-673,1795), beacon (375,1037,1501), beacon 3 (-93,1836,1888). The noise souce is located at (600,800,1800). Fig. 11. Flowchat of NRW algoithm (Fig. 11) is flowchat fo NRW algoithm, whee the localize eceives ultasonic signals fom beacons sequentially. Using the ultasonic flight time, the flight distance is calculated. With the calculated distance, the filteing pocess checks whethe the data exist within the toleance bound o not. Unless the data exist within the toleance bound, it is consideed as a malicious signal affected by obstacles o envionmental noises. In this case, the cuent flight distance is estimated as d ˆ Bij, the summation of the pevious flight distance and the standad deviation of the flight distance in the window as shown in (13),. In the othe case, the flight distance is assumed to be eliable and it is selected as d ij, the updated flight distance fo the cuent step. When the estimated values of d ˆ Bij ae used continuously, the absolute eo becomes lage and lage. This situation is handled by efeshing the window when the estimated values of d ae ˆij used moe than fou times consecutively. The numbe fou is heuistically and empiically detemined with the window size of twenty. 5. Expeiments and Results Fo indoo localization of the mobile obot to veify the effectiveness of the newly poposed NRW algoithm, a mobile obot is designed with the installation of igs. The igs consists of localize and beacon. The localize which is caied by the mobile obot sends out RFID signals and eceives ultasonic signals to measue the flight time of the ultasonic signal; the beacon eceives the RFID signal and checks with its own ID to send out the ultasonic signal when the two codes match to each othe. In the localize, fpic1f675f is used to contol the RFID, TMS30C06A is used fo system contol, and AR0-10P(0kHz)is used to eceive the ultasonic signal. In the beacon, MSP03F1101A is used fo the system contol, and AT0-10P(0kHz) is used fo the ultasonic tansmission. Noise souces ae placed in the expeimental envionment without any specific ID. The mobile obot consists of diving pat and contol pat. Fo the diving pat, S56Q-0355S stepping motos ae utilized to move and PIC18F870 is used to contol the motos. The maximum velocity of the mobile obot is kept as 35cm/s not to cause dangeous situations to human. Fo the expeimental Fig. 1. Expeimental envionment Fig. 13. Tajectoy of obot unde the noise envionment The noise chaacteistics should be andom and unexpectable. Howeve the continuous noises which have the same fequency as the beacon cannot be filteed out. Theefoe in this expeiment the noise souce is activated evey 00 ms while the mobile obot is moving along a cicula path as shown in (Fig. 13). The small cicles in (Fig. 13) epesent the measued locations of the mobile obot while the mobile obot is moving on a cicula path maked as X. By the noise intefeences, the localization pocess is not successful at all. Fig. 1. Without NRW algoithm 76

6 17th IFAC Wold Congess (IFAC'08) Seoul, Koea, July 6-11, 008 (Fig. 1) epesents the measued tajectoy of the mobile obot along X-axis and Y-axis espectively, while it is moving on a cicle unde noisy envionment. ultasonic sensos and to povide a cheap and obust localization system fo a mobile obot moving in a wide indoo envionment. The auto calibation algoithm shotens localization time and impoves the data eliability when the mobile obot moves into a new block. The NRW algoithm povides a obust localization scheme to eliminate the noises caused by the obstacles in the ultasonic signal path as well as the active envionment itself. Effectiveness of the poposed algoithm is demonstated though the eal expeiments. 7. Acknowledgements "This wok was patly suppoted by the IT R&D pogam of MIC/IITA[005-S-111-0, intelligent obot senso] and Ninety System co., Ltd". Fig. 15. Tajectoy of obot with NRW unde the noise envionment To show the effectiveness of NRW algoithm, the same expeiment is pefomed by applying NRW algoithm fo the cicula motion of the mobile obot. The esults ae illustated in (Fig. 15). It is clea that with NRW algoithm, the mobile obot caying igs ecognizes eal locations pecisely while thee ae seveal intefeences of ultasonic noise souces. Fig. 16. With NRW algoithm (Fig. 16) epesents the obot tajectoy in tems of X-axis and Y-axis espectively, while thee exist ultasonic noise souces. 6. Conclusions To build ubiquitous envionment fo a mobile obot moving inside buildings o undegound whee GPS signals cannot get into clealy, igs using ultasonic sensos povides a localization system which is cheap, has fast data update ate, and povides high position accuacy. Fo the outdoo envionment, GPS is popula. Howeve the ultasonic signals ae vey sensitive to noises. Theefoe ultasonic sensos ae malicious when thee ae some obstacles in the signal path o when the envionment itself is noisy. This pape poposes a block expansion algoithm with an auto calibation and NRW algoithms to ovecome these shotcomings in using REFERENCES San Mok Lee (001). Using Vision Senso and Sona Senso of Mobile Robot Localization. Chungnam National Univesity. H. Pemans and V. Campenhout (1993). High Resolution Senso Based on Ti-aual Peception. IEEE Tans. On Robotics and Automation, vol. 9, no.1, pp A. M. Flynn (1998). Combining Sona and Infaed Sensos fo Mobile Robot Navigation. The Intenational Jounal of Robot Reseach, vol. 7, no.6, pp Tatsuo Aail, and Eiji Nakatno (1983). Development of Measuing Equipment fo Location and Diection Using Ultasonic Waves. ASME Tans., vol. 105, pp S. Hong, M. H. Lee, S. H. Kwon, and H. H. Chun (00). A Cat Test fo the Estimation of GPS/INS Alignment Eos. IEEE Tans. Intelligent Tanspotation Systems, vol. 5, no.3. Billu Bashan (000). Fast Pocessing Techniques fo Accuate Ultasonic Range Measuements. IOP J. Meas. Sci. Technolog, vol. 11, pp M. Paialla, J. J. Anaya and C. Fitsch (1991). Digital Signal Pocessing Techniques fo High Accuacy Ultasonic Range Measuements. IEEE Tans. Instumentation and Measuement, vol. 0, no., pp S. B. Kim and JangMyung Lee (007). Mobile Robot Localization Using Auto Calibation Algoithm. KRS Jounal, vol., no. 1, pp S. B. Kim and J. M. Lee (006). Robust Positioning a Mobile Robot with Active Beacon Sensos. LNAI 5, ISSN , Pat 1, pp S. B. Kim, I. O. Lee and J. M. Lee (006). Dynamic Localization of a Mobile Robot with Active Beacon Sensos. th IFAC Symposium on Mechatonic Systems Heidelbeg Gemany, Sept Iowa State Univesity GPS page Web site at 77