Internatonal Journal of Materals, Mechancs an Manufacturng, Vol. 1, o. 4, ovember 2013 The Investgaton of the Obstacle Avoance for Moble Robot Base on the Mult Sensor Informaton Fuson technology Qu Dongyue, Hu uanhang, an Zhang utng II. Abstract In orer to accurately bul a moel about external envronment of moble robot, a robot mult-sensor system s esgne. Ultrasonc sensor array aopt a strbute arch-tecture, nfrare stance sensor use the esgn of class laser raar type, usng mult-sensor nformaton fuson technology, establsh a 180 scrollng wnow n front of the robot, though the metho of polar coornates vector to obstacle etecton an external envronment moel, usng the rollng wnow obstacle avoance algorthm, robots acheve autonomous obstacle avoance an navgaton n unknown envronment. The expermental results confrme that the robot system an the valty of the theory mentone n ths paper s val an optmal. In ths stuy, to realze the autonomous robot control an navgaton, the man control center of the esgne autonomous navgaton robot s a personal computer IPC, whch uses the C++ language to wrte the software program by conserng the stablty, ynamcs an optmzaton. The robot has four parts: 1) Energy layer. The layer s powere by a 12V, 2400mA lthum battery-powere. 2) Drver layer. Moble robot by Sabertooth ual 10A DC motor rve proves up to 10A of contnuous rvng, meanwhle usng four 12V Coreless planetary gear motor. 3) Sensor layer. The sensor layer has sx ultrasonc sensors, an nfrare stance sensor, an electronc compass sensor an an oometer. 4) Master layer. The master layer s compose of 1 personal computer PC, whch can complete the acquston of sensor nformaton, mult-sensor nformaton fuson, the plannng algorthm processng an other functons. The autonomous navgaton robot n have the followng characterstcs: The acton of the robot s flexble an relable an the movement lnes are smooth an contnuous:the overall structure of the robot s taken by moular esgn, whch s easy to control an acheve. Inex Terms Moble robot, communcaton fuson, polar coornate vector, rollng wnow. I. ITRODUCTIO Sensor technology s wely apple n the fel of ntellgent robotcs research [1].There s easly nterference by the envronment n sngle sensor whch causes large errors an the lmte etecton range [2]. Mult-sensor nformaton fuson technology can acheve the goals of beng complementary avantages of sensors an reuce reunant nformaton; meanwhle t can mprove the ecson-makng an respon ablty of the system. So mult-sensor nformaton fuson technology has broa applcaton prospect [3]. At the present stage, mult-sensor nformaton fuson technology manly reles on the theoretcal algorthm, such as Kalman flterng metho, Bayesan metho, least squares metho[4], but less for real applcatons. Base on the practcalty, an autonomous navgaton robot s evelope. To bul an accurate moel of the external envronment, the robot mult-sensor system wth ultrasonc sensors, nfrare stance sensor, electronc comp-ass sensor an oometers esgne. Ultrasonc array has the strbute archtecture, an the nfrare stance sensor has the class laser raar type esgn, whch realzes sensor mult-speces, the number of nformaton fuson. The expermental results emonstrate the effectveness an practcalty of the system esgne [5]. III. MULTI-SESOR IFORMATIO FUSIO SSTEM Mult-sensor nformaton fuson technology s an effectve nformaton-processng technology, whch has a mult-sensor type, number of collaboraton achevng complementary avantages of the sensor, hgh system relablty an strong ant-nterference ablty. The moble robot selects ultrasonc sensors an nfrare range sensors as stance etector.the ultrasonc sensor has large measurng range, but t has a cross-talk feature an s easy to prouce phantom phenomenon, whch coul easly cause large accental errors [6]. Infrare stance sensor can compensate the ultrasonc sensors for the bln spots, but t has the smaller measurng range. So the ultrasonc sensors an nfrare range sensors exchange nformaton to realzng complementary avantages an a more accurate external envronment moel [7]. A. Ultrasonc Sensors Dstrbute System To the goal of autonomous obstacle avoance an navgaton, the moble robot nees comprehensvely survey of the external envronment. External envronment of robot s ve nto two parts: 180 front area an 180 back area. In the process of robot autonomous navgaton, the front Manuscrpt receve December 28, 2012; revse February 2, 2013. Sources of project: the freeom to explore of central unversty research busness (HEUCF100709). Qu Dongyue an Hu uanhang are now wth the Harbn Engneerng Unversty, Chna (e-mal: quongyue@hrbeu.eu.cn, huyuanhanghgc@163.com). Zhang utng s wth the Insttute of Avance Technology of Helongjang Acaemy of Scence, Chna. DOI: 10.7763/IJMMM.2013.V1.79 AUTOOMOUS AVIGATIO ROBOT SSTEM DESIG 366
Internatonal Journal of Materals, Mechancs an Manufacturng, Vol. 1, o. 4, ovember 2013 area of 180 s nevtable etecton area. In ths paper, sx ultrasonc sensors are use to etect the front of the robot 180 area. If concave trap an local extremum are encountere, robot wll spn aroun 180 an change ts front an rear, then the new path wll be planne agan. Ultrasonc sensors are etecte n a strbute nstallaton metho an every sensor has 60 cone angle, Rght horzontal lne as a reference lne, the face perpencular to the center lne of ultrasonc sensors B, C, D, E, F, G an reference lne sequentally nto a 30, 60, 90, 90, 120, 150 nstallaton, the sensor arrangement shown n Fg. 1: Fg. 1. Arrangement of ultrasonc sensors an jugment about azmuth of obstacles. In Fg. 1, each sensor emts two rays representaton of etecte regon bounary lnes, short black ots ncate the obstacle. The angles n the chart are the azmuth angle of the obstacle that entfe by two ajacent sensors. The etecton range of optonal ultrasonc sensor s 4cm 160cm. The black trangle area s bln spots of etecton about two ajacent ultrasonc sensors, an the area s etecte an ae by Infrare rangng sensors. Base on the ultrasonc sensor array obstacle azmuth, the practce s shown below: The stance values measure by the sx ultrasonc sensors are represente by L (1 6), whch s n the range of 4cm-160cm. It shows a maxmum 160 when the ultrasonc sensor cannot etect obstacles, the actual stance between the obstacles to the robot s represente by L, the angle of the obstacle s represente by. The angles of 6 ultrasonc sensors arrange on the robot are known, usng the number of the two ajacent ultrasonc sensors to show the obstacle angle nformaton, just as the pcture 1 shown. Establsh the basc rules lbrares as follows: : If P then Q ncates -th rule, P represents the contons of the rule, Q represents the results of the rules. 1) If 4 L1 160 an L2 L L1 an 30 ; L 2) If 4 L1 160 an 4 L2 mn( L, L ) an 45 ; 1 2 3) If 4 L2 160 an L3 L L2 an 60 ; L L L 4) If 4 L2 160 an 4 L3 mn( L, L ) an 75 ; 2 3 5) If 4 L3 160 an 4 L4 mn( L, L ) an 90 ; 3 4 6) If 4 L4 160 an 4 L5 mn( L, L ) an 105 ; 4 5 7) If 4 L5 160 an L4 L L5 an 120 ; L 8) If 4 L5 160 an 4 L6 mn( L, L ) an 135 ; 5 6 9) If 4 L6 160 an L5 L L6 an 150. The subject selecte sx ultrasonc sensors to arrange whch can approxmately get the nformaton of nne azmuths. It s sure that you can set more ultrasonc sensors. Wth the ncreasng number of sensors, the accuracy of jugng the obstacle azmuth wll ncrease. B. Establsh Obstacle Moel wth Polar Coornate Vectors Establshng obstacle moel wth polar coornates vector metho s an mprovement of gr metho[8]. The prncple s to store the external characterstcs of the obstacle n the form of recton vector, but only to o ths. It overcomes the shortcomngs of the tratonal gr metho, such as large reserves an long cycle moelng, etc. The metho to create a polar coornate vector s as follows: frstly, ve the 180 rollng wnow nto n, an then establsh the vector wnow mae up of 180 ( n 1) vector lne to establsh obstacle feature moel by the vector lnes. Of course, the greater the value of n, the more ntensve the vector s an the more accurate the moel s, but the amount of ata that nee to be hanle also ncreases. The mert of establshng s usng a lne nstea of a bunch of gr [8]. An the selecton of n shoul follow the prncple that each gr s mappe at least one vector lne. δ shoul satsfy the followng equaton: 180 ( n 1) 2R 2R cos 2S 2 2 2 s the angle two ajacent vector lnes. R s the threshol of scrollng wnow.s s the length of the gr se. The gr sze n ths subject s 16 16cm, the threshol of scrollng wnow s 60cm.Havng calculate, can take as18, so t can create 11 recton vectors. ow we assume the length of th vector as, f R, the vector wll be an effectve vector. The start angle of rollng wnow s 0, step angle s 18, the recton of the vector angles are etermne as follows: 1 0, 2 18, 3 36 11 180, an ultmately we create a recton vector L(, ) collecton as the obstacles contour curves, the range of s 1 11. Of course, the number of vector lne n, envronmental moelng more accurately, the corresponng nformaton processng capacty s greater. Polar coornates vector metho to establsh the prncple of the obstacle moel shown n Fg. 2: Fg. 2. The establshment of the polar coornates Vector 367
Internatonal Journal of Materals, Mechancs an Manufacturng, Vol. 1, o. 4, ovember 2013 The avantage of establshng obstacle moel wth polar coornates vector metho s that accurate moel, small reserves, fast response of robot. C. Desgn of Infrare Rangng Sensor a Class Laser Raar Establshng obstacle moel wth polar coornates vector metho requres nfrare rangng sensor system to create a 180 rollng wnow n front of the robot an smultaneously etect the stance an azmuth angle of the obstacle. To the above requrement, a class laser raar system base on nfrare rangng sensor s esgne. Specfcally as follows: nstall a nfrare rangng sensor to steerng gear hea, etect by the nfrare rangng sensor, get feeback angle by steerng gear hea. Eventually real-tme etecton of parameters of obstacle an azmuth are acheve. Infrare stance sensor system nstalle poston of pont A n Fg. 2. The class laser raar system conssts of three parts: 1) pan-tlt 2) nfrare rangng sensor 3) the man control center. The nfrare rangng sensor s mounte to the 180 rotaton of the steerng gear hea, an the steerng gear rotaton s accurately controlle by USBSS32 control panel.usbss32 goes on ata communcatons va USB nterface wth PC, whch can mplement the settng of the scan ntalzaton, stop scannng, etc. The man control center uses a PC. Each tme t eals wth the stance nformaton measure by the rotaton angle an the nfrare rangng sensor of the steerng gear PC as a smultaneous acquston an storage, after analyss an processng of ata, the external envronment nformaton s efne. There are ten pulse angle of the steerng gear n ths stuy, whch can complete exploraton msson well. Wth the ncreasng number of servos pulse wth, the number of rollng wnow polar coornates vector lne n also ncreases, an the angular resoluton s hgher. But the probe cycle wll be longer. Ths shoul be etermne accorng to the actual stuaton. Accorng to the above methos, establshng a class laser raar system wth the applcaton of nfrare rangng sensor an bulng the 180 rollng wnow n front of the robot can accurately establsh the obstacle envronment moel n polar coornates vector metho [10]. D. Dve of Mult-Sensor Informaton Fuson Level The strategy of moble robot moton s ve nto four levels: the layer that ten to target; reucton layer; the layer of accurate response; the layer of emergency response. The functon of each level s as follows: The layer that ten to target: When Sensor system oes not etect obstacles, the robot executes Strateges ten to target. Reucton layer: When the ultrasonc probe area etect obstacle frst, the robot performs reucton strategy to prepare for the mplementaton of the follow-up obstacle avoance strategy. The layer of accurate response: When obstacle s etecte by the Common etecton area, the robot performs accurate obstacle avoance strategy. The layer s the key one to the robot obstacle avoance system. The layer of Emergency response: When n an emergency, such as the robot encounters concave trap, local extreme, the robot can not avo obstacles n precse plannng layer, then the robot perform emergency obstacle avoance strategy. the layer that ten to target reucton 减速慢行层 layer 精确规划层 紧急规划层 layer of accurate response layer of emergency response area of the ultrasonc probe area of common etecton area of nfrare etecton Fg. 3. Herarchcal vson of mult-sensor nformaton fuson In the four levels, the layer of accurate response s the core part of the levels. So the followng s the stuyng about the scopng of the accurate response layers. The layout of Sensor s shown n Fgure 1. A s the nfrare stance sensor; every one of B, C, D, E, F an G s an ultrasonc sensor, an they are putte n symmetrcal arrangement. H, I, J are the vertces of the etectng bln spot of the two ajacent sensor. Calculate AH, AL, AJ an choose the maxmum value as the mnmum raus for bln spot etecton complemente by nfrare rangng sensor an the efne lne whch s the common etecton area of ultrasonc sensor an nfrare rangng sensor. Known: The length of robot s 36cm, an wth s 30cm. A, B, G are collnear, an they has a 8cm stance from the bottom lne. D an E are collnear an a 4cm stance from the top lne. AB=15cm, BC=16cm, DE=14cm. CBH 30, JDE 60 Through mathematcal calculatons: AH=37.35cm, AI=40.71cm, AJ=36.12cm. So select the maxmum value of 40.71cm for startng bounary of the common etecton area. In ths paper, the etecton range of the Infrare Rangng sensor s from 10cm to 80cm. The etecton range of the ultrasonc sensor s from 4cm to 160cm. Infrare stance sensor n nstallaton poston of A as the reference. Wth the above analyss, we can etermne the range of the nfrare etecton that from 10cm to 40.71cm. The structural mensons of the robot just to make up Infrare rangng sensors etect bln spots. A large number of experments confrme that: 80cm s most approprate to the vson of reucton layer an the layer of accurate response shoul. So the area of Common etecton s 40.17cm~80cm, the area of the ultrasonc probe s 80cm~160cm. The pont of nfrare rangng sensors locate as a reference. Every level s as follows conserng the robot mensonal boywork: The layer of Emergency response: <40.71cm; The layer of Accurate response: 40.71cm~80cm; Reucton layer: 80cm~160cm; The layer that Ten to target: >160cm. Experments show that the esgn of moble robot's varous levels makes the robot more stable performance, an the etecton result wll be more accurate IV. OBSTACLE DETECTIO AD EXTERAL EVIROMET MODELIG Moble robot establshes rollng wnow, then obstacles polar agram s set up by the contrast obstacle etecton stance an the valve value R of rollng wnow. Base on 368
Internatonal Journal of Materals, Mechancs an Manufacturng, Vol. 1, o. 4, ovember 2013 the analyss an processng of polar coornates vector graph ata, the obstacles of the characterstc parameters are etermne nclung obstacles startng angle, en angle an the mnmum stance of the obstacle to robot reference pont A. All the nformaton above can etermne the basc nformaton of an obstacle. In the etecton process of robot obstacles, frstly the concept of bounng box s ntrouce. That s: In moble robot moton process, when foun that obstacles mplement obstacles avoance strategy, the obstacle moel shoul be expane treatment at frst to prevent collson between the robot an obstacles ege. The expane vrtual obstacle moel s calle the bounng box, an the stance between the ege an the orgnal obstacles s.the bgger s, the hgher obstacle avoance safety coeffcent s. When R, rollng wnow wll be consere that there are obstacles exst nse n the process of etectng obstacle. The scrmnaton of obstacles types n robot system can be ve nto fve kns as follows: 1) When R, 1 R an 1 L, the system etermnes the same obstacle an the etecton contnues. 2) When R, 1 R an 1 L, the system etermnes two obstacles overlappng. 3) When Ran 1 R, the system etermnes that a obstacle etecton has ene. 4) When R an 1 R, the system etermnes that a new obstacle etecton starts. 5) When R an 1 R, the system etermnes that there are no obstacle aroun the robot. Among them, s the length of artcle vector lne n rollng wnow, R s a threshol of rollng wnow. L s the safety through the spacng of robot, L D 2, D s the wth of the robot boy, s the puffe stance of obstacle. Moble robot usng polar coornates vector metho to establsh moels of obstacle process as shown n fgure 5: Autonomous navgaton robot through the rollng wnow to establsh external envronment moel, an can acheve obstacle etecton an obstacle overlap the complcate stuaton of jugment, then we can well establsh external envronment moel, for the robot autonomous obstacleavong an navgaton reay. R V. EXPERIMETS ABOUT OBSTACLE AVOIDACE OF THE AUTOOMOUS AVIGATIO ROBOT Ths paper esgne autonomous moble robot, establshe the ultrasonc rangng sensor system, nfrare stance sensor laser raar system, usng mult sensor nformaton fuson technology, combne wth the polar coornate vector metho to establsh the envronmental moel, robot obstacle avoance algorthm apple to scroll the wnow of obstacle avoance an navgaton. The followng experment s use MATLAB smulaton an robot platform to verfy the valty of the propose theory. In the MATLAB smulaton envronment, the moel shown n Fgure 5: n the fgure, the re represents the moble robot, n front of the re establsh a 180 scrollng wnow, the corror s the both ses of the black part, the black mle part s the obstacle moel Fg. 5. Smulaton about autonomous navgaton From the fgure can be seen, the moble robot has complete the path selecton from the start pont to the target pont, an successfully avoe the obstacle encountere by the avance, an the robot path selecte s feasble an optmalty. Confrmng the effectveness of mentone algorthm[9] In the research, we use the platform of autonomous moble robot to o obstacle avoance experment. In the experment, we use three cartons an one fnshng tank as the obstacles. the export on the left of fnshng box s the target poston of the robot. Fgure 6 s the process of autonomous moble robot autonomous navgaton. start Establsh rollng wnow j=r no obstacle have obstacle j+1=r j+1=r no obstacle new obstacle j+1-j L en of the obstacle overlap the obstacles regar as one obstacle fnsh scannng en Fg. 4. Detectng obstacles an moelng flow chart Fg. 6. Obstacle avoance experments wth many obstacles Robot urng exercse, ultrasoun etecton zone frst etects obstacles, robot n the eceleraton chronc moe. The robot move on, enter the exact plannng layer, startng a real-tme obstacle avoance strategy, the robot avo obstacles encountere, eventually reachng the target pont. As can be seen from the fgure, the robot nepenently choose the path of feasblty, optmzaton, an confrme 369
Internatonal Journal of Materals, Mechancs an Manufacturng, Vol. 1, o. 4, ovember 2013 the excellent characterstcs of the propose metho. VI. COCLUSIO An autonomous moble robot, compose of energy layer, rver, sensor layer an control layer, was esgne n ths paper. The ultrasonc rangng sensor system an nfrare stance sensor laser raar system were establshe n the robot. The technology of mult-sensor nformaton fuson was apple. 180 rollng wnow was bult n front of the robot, whch coul be ve nto the tren of the target layer, slow own layer, precse plannng layer an emergency plannng layer. Combne wth polar coornates vector metho for obstacle etecton an the external envronment moel, the robot can precsely etect the envronment. The algorthm of the rollng wnow obstacle avoance was employe to realze the autonomous obstacle avoance an navgaton of the robot n an unknown envronment. The effectveness an the optmzaton of the robot esgne system an the presente theory were valate by the results of the experments n ths project. [6] H. Boubertakh, M. Tajne, an P.. Glorennec, A new moble robot navgaton metho usng fuzzy logc an a mofe Q-learnng algorthm, Journal of Intellgent Fuzzy Systems, vol. 21, pp. 113-119, 2010. [7] F. u, Research on the Mult-moe Control System of Intellgent Wheelchar, anchang: anchang Unversty, 2009. [8] B. Sun, D. Han, an Q. We, Applcaton of Rollng Wnow Algorthm to the Robot Path Plannng, Computer Smulaton, vol. 23, no. 06, pp. 159 162, 2006. [9] J. L. Blanco an J. Gonzale, an J. A. Fernanez Marga, Extenng obstacle avoance methos through multple parameter space transformatons, Autonomous Robots, vol. 24, no. 01, pp. 29-48, 2008. Qu Dongyue was born n 1973, who grauate from the Moscow State the Process Unversty STAKI n 2008. He has obtane hs Ph.D egree. He s now workng at Harbn Engneerng Unversty. The man research areas are moble robot control technology Hu uanhang was born n 1986. He s now stuyng at the Harbn Engneerng Unversty, hs current research areas aremoble robot sensng an control technology. REFERECES [1] L. Ren, Obstacle Percepton an Obstacle-avong Strategy Pesearch of Moble Robot Base On Laser Range Fner, Harbn: Harbn Insttute of Technology, 2007. [2] H. A. Hagras, A Herarchcal Type22 Fuzzy Logc Cont rol Archtecture for Autonomous Moble Robot, IEEE Trans. actons on Fuzzy Systems, vol. 12, no. 4, pp. 524-539, August 2004. [3] Z. J, Moble Robot Target Detecton an Target Recognton Fuson Technology Research, Jnan: Shanong Unversty, 2011. [4] J. Ba, L. Chen, H. Jn, R. Chen, an H. Mao, A new path plannng for robot n ynamc an unknown envronments, Transucer an Mcrosystem Technologes, vol. 30, no. 10, pp. 33 36, 2011. [5] S. Km an. Km, Robot Localzaton Usng Ultrasonc Sensors, n Proc. IEEE/SRJ Internatonal Conference on Intellgent Robots an Systems, Sena, Japan, 2004, vol. 4, pp. 3762-3766. Zhang utng was born n 1976. She grauate from the Moscow the State Process Unversty STAKI n 2006. Her current research areas are moern nustral technology. 370