Telecontrol of AutoMerlin Robot by Employing Fuzzy Logic

Transcription

1 Internatonal Journal of Mechancal Engneerng and Robotcs Research Vol. 5, No. 1, January 2016 Telecontrol of AutoMerln Robot by Employng Fuzzy Logc Aamr Shahzad, Ram Al-Jarrah, and Hubert Roth Insttute of Automatc Control Engneerng (RST), Segen, Germany Emal: {aamr.shahzad, ram.al-jarrah, perform dfferent tasks by human operator from control staton whch s stuated at a safe place from operaton ste. In present day world, they fnd ther effectveness n explosve dsposal, landmnes clearance, nuclear waste handlng, planet exploraton, complcated ndustral constructons et al [3]-[5]. Teleoperaton of moble robot s also beng employed n medcal feld for treatment of patents usng the technque of tele-surgery. They are becomng the essental part of hosptals for performng a varety of tasks [6]. Teleoperated robots are beng utlzed n educatonal nsttutes for tranng and learnng purposes. They are also beng used n servce ndustry for gudng humans n bg malls and flyng robots teleoperaton s helpng n montorng large gatherng of peoples by mountng vdeo cameras on them. Teleoperaton has become more and more affordable usng Internet as communcaton medum because t s cheap and s easly avalable these days. It has fast transmsson rate and therefore t s prmary choce of researchers to utlze ts avalablty and benefts to teleoperate robots. Therefore, nowadays Internet based teleoperaton of moble robots has opened new opportuntes n long dstance learnng, resources sharng, and remote expermentaton [7]-[9]. Ths paper s structured as follows. Secton II s related work n the area of telecontrol usng dfferent approaches have been elaborated wth a few examples from recent research. In Sectons problem descrpton has been elaborated, t s descrbng the workng of control loop and challenges nvolved. Secton IV deals wth the Fuzzy control approach and ts mplementaton n current scenaro. Sectons V shows the expermental results and ther descrpton and Secton VI concludes the paper wth a bref outlne on future work. Abstract In ths work the novel teleoperaton of real moble robot s presented. In order to cope wth the ssue of random tme delay whch makes system unstable, Fuzzy logc approach has been used to desgn a stable controller to teleoperate AutoMerln moble robot. The fuzzy controller has been desgned n such a way that t takes two nputs to gve one output. Frst nput s generated by human operator from control staton usng master devce lke joystck. The second nput to the controller s generated based on the obstacles n front of moble robot. The onboard proxmty sensors scan the envronment and the algorthm deployed on the moble robot as local ntellgence determnes the locaton of nearest obstacle and feeds ts value to controller as second nput. When controller gets both nputs then t calculates the approprate speed for moble robot. The fuzzy rules have been defned n such a way that they protect the robot from colldng wth obstacles and random delay does not affect the performance and stablty. The presented results exhbt the performance and stablty of overall system n the presence of random tme delay. Index Terms telecontrol, ground robot, fuzzy logc, fuzzy sets model, teleoperaton, tme delay I. INTRODUCTION The man ambton of a teleoperaton system s to manpulate two separate systems and to make the remote system track closely the local system. The teleoperaton system s composed of a human operator, master devce, local control staton, communcaton channel and slave moble robot. The master devce s maneuvered by a human operator and a slave moble robot s desgned to follow the human operator's commands [1]. Teleoperaton s prmary choce to control moble robots navgaton n unknown, unstructured and dangerous envronments. The presence of human operator n teleoperaton s benefcal due to hs/her sophstcated cogntve capabltes [2]. As compared to autonomous robots, teleoperated robots can utlze human wsdom to complete the desred navgaton, manpulaton or handlng dangerous materal task more accurately because human operator can take dfferent decsons accordng to envronment and stuaton demand. Another advantage of teleoperaton s safety of human operator because the envronment n whch teleoperated moble robot operate s often dangerous for human health. Therefore, teleoperated moble robots are employed to II. The moble robot teleoperaton has been dscussed wth vsuo-haptc nterface [10]. The man focus was to drve the robot n unknown maze-lke rectangular envronment. However they dd not consder tme delay nvolved n telecontrol of moble robot. As there s always some delay n teleoperaton whch makes the system unstable, ths crucal problem has been left for future work. Moble robots prototype has been developed for teleoperaton usng RF transcevers [11]. The man dea behnd development of such prototype s survellance. But t dd not talk about stablty, tme delay and dffcultes due to unstructured envronment. In Ref. Manuscrpt receved July 1, 2015; revsed October 16, do: /jmerr RELATED WORK 17

2 Internatonal Journal of Mechancal Engneerng and Robotcs Research Vol. 5, No. 1, January 2016 [12] sxteen ultrasonc sensors are attached around a moble robot n a rng pattern to measure the dstances to the obstacles and a collson vector s ntroduced. A vrtual reflecton force s generated to avod the obstacles and then the reflecton force s transferred to the operator who s holdng the joystck whch s used to control the moble robot. Based on ths reflecton force, the operator can control the moble robot more smoothly and safely. The ncorporaton of motor nonlneartes and human hand dynamcs n order to convey the envronmental nformaton to the operator more realstcally s left as a future work, whch s necessary for the moble robot to navgate n the envronment wth random obstacles on the rregular surface. Many researchers have used passvty control and scatterng theory approach to desgn sutable controller for teleoperaton of moble robot [13], [14]. However, they have evaluated the system based on pretty smple tasks by gnorng the crucal problem of presence of tme delay n teleoperaton and further expermentaton and evaluaton of system n complex envronments ncludng the presence of communcaton delay n the data transmsson s under plannng [13]. Ref. [14] the work s based on smulaton and t has not taken nto account the unstructured envronment whch s always present around the slave robot. Therefore, t s hard to mplement n real world. It s mpossble to teleoperate the robot wthout havng any knowledge about envronment around the robot. Ref. [15] a desgn of the telecontrol electrcal vehcle system s presented usng wave varables. Contrary to real world random delay, n ths smulaton work tme delay s assumed to be constant and envronment s assumed to be passve. The human operator needs accurate percepton of envronment around the robot to teleoperate the robot effcently. Otherwse, t would be tedous task for hm to teleoperate the robot. Therefore, n order to dexterously teleoperate the robot we have presented a fuzzy logc controller for the navgaton of robot n ll-structured envronment by consderng obstacles n the envronment around the robot. III. PROBLEM DESCRIPTION The telecontrol setup s shown n Fg. 1. It conssts of a joystck whch s connected to a laptop. The laptop hosts the algorthm whch translates the joystck movement nto nput velocty for robot. It also dsplays the vsual feedback from the robot. Ths whole setup s regarded as control staton. The human operator stays n control staton and sets dfferent commands for the robot navgaton at remote locaton. The velocty command travels from control staton to the robot through communcaton channel. The communcaton has been establshed usng TCP/IP protocol. So at control staton human operator acts lke clent and on the other end of communcaton channel there s a mcro pc whch controls the behavor of robot acts as server. When the velocty command reaches the server pc t sends t to the robot va seral port and receves vsual and sensory feedback from robot. The telecontrol suffers from tme delay due to communcaton between control staton and robot. Ths delay depends upon the dstance between the control staton and the moble robot. It has fxed value lke when the dstance s fxed and there s dedcated medum of communcaton between the control staton and the robot lke planetary exploraton or t has random value lke communcaton through Internet. Ths delay s vtal for teleoperaton of moble robot. Ths delay causes the system to become unstable. Another ssue s correct presentaton of envronment to human operator so that he can take approprate decsons. Among the sensors used for envronment recognton, the camera s the most popular and powerful. However, there are several lmtatons f only camera s used n teleoperaton of a moble robot. For example, shadows and curved areas cannot be vewed usng a narrow vew angle camera, especally n an envronment wth bad llumnaton and several obstacles. Therefore, n ths work we have mplemented the fuzzy logc to desgn a sutable controller whch can work n llstructured envronment potentally dangerous to human health, n the presence of random communcaton delay. The desgn s based on two nputs. One nput comes from the human operator and other from an algorthm developed and deployed on the robot, whch takes the values of onboard proxmty sensors and sort out the shortest one. Ths shortest dstance value from robot to obstacle s fed as second nput to fuzzy controller gve approprate output speed. The mplcaton of fuzzy logc enables us to teleoperate the AutoMerln robot n the presence of random tme delay. The robot lstens to human operator and follows hm/her untl there s no obstacle n front of hm. But f t detects the obstacle then t calculates the speed based on the objects n front of t and slows down to zero speed f the robot reaches very near to object. Then robot can lsten to operator for reverse movement only but gnores the commands for advancement. Addtonally, the lve vdeo feed s made avalable to operator so that he can perceve the envronment. IV. INTELLIGENT FUY SETS MODEL The mostly used controller for the nonlnear system s the fuzzy controller. The fuzzy system s able to have robust control of the robot and t has low senstvty to a varaton of parameters or nose levels [16]. As t s known, the fuzzy logc s a powerful problem-solvng methodology wth a myrad of applcatons n embedded control and nformaton processng. It could provde a remarkably smple way to draw defnte conclusons from vague nformaton. In fuzzy reasonng, the most mportant fuzzy mplcaton nference rule s the Generalzed Modus Ponens (GMP), whch uses IF-THEN rules that mplctly represent a fuzzy relaton. The fuzzy rules defne the connecton between nput and output fuzzy (lngustc) varables. The fuzzy rule based system s composed of four man stages: Fuzzfer, Rule base, Inference engne, and Defuzzfer [17]-[19]. The fuzzfcaton stage s defned as the process to convert the 18

3 Internatonal Journal of Mechancal Engneerng and Robotcs Research Vol. 5, No. 1, January 2016 crsp value nput u0 to a fuzzy pont A' whch s defned n (1). w* (w )w (w ) c* A 1, (u ) 0, f u u (1) otherwse In the rule-base system for more than one rule, the fuzzy rules that descrbe the fuzzy system are gven as the followng: r1: IF Joystck s Hgh Negatve AND Obstacle Dstance s Close THEN Speed s Low Negatve... rk: IF Joystck s Low Postve AND Obstacle Dstance s Far THEN Speed s Low Postve... rn: IF Joystck s Hgh postve AND Obstacle Dstance s Very Far THEN Speed s Hgh Postve Therefore, the translaton of ths rule base nto a fuzzy relaton s done by constructng the fuzzy relaton Rn for each rule rk; and then combnes these relatons nto sngle fuzzy relaton R as descrbed n (2). Ths process for combnng the fuzzy rules nto a fuzzy relaton s called aggregaton. Ths aggregaton reles on the type of the relaton used to represent the rule n the rule base system. In case of Mamdan approach Rk s represented by usng conjuncton operator; then the aggregaton s done by usng dsjuncton operator [20]. R Rk c * (4) The ntellgent fuzzy sets model s dvded nto two man parts. Frst, there are three sonar sensors (S1, S2, S3) mounted n front of the robot to detect obstacles n the envronment. Based on the probablty and the angle between the sensors whch s here ϕ= 20, the algorthm wll fnd the shortest dstance between the robot and any obstacle based on these three readngs as t s shown n Fg. 2. Here we have assumed that the vew angle of the sonar sensor s So there s no conflcton between the sensor readngs. Accordngly, there are three probabltes for the sensors: S1 s the shortest, S3 s the shortest or S2 s the shortest. Second, fuzzy control was desgned n order to control the speed of the robot and deals wth random delays due to communcaton through shared medum lke Internet. The controller s based on two nput varables and one output. The frst nput s the joystck command that comes from the user wth fve membershp functons and the lngustc varables are (hgh negatve HN, low negatve LN, zero, low postve LP, and hgh postve HP) as t s llustrated n Fg. 3. The second nput comes from algorthm. It has 5 trangular membershp functons and 5 lngustc varables (very close VC, close C, medum M, far F and very far VF) as t s shown n Fg. 4. The controller has one output that s the speed of the robot wth fve trangular membershp functons (negatve hgh NH, negatve low, ero, postve low, and postve hgh PH) as shown n Fg. 5. The generalzed modus ponens and T-norms have been used to desgn the fuzzy rules base wth 25 rules n terms of relatonshp between nputs and output as t s llustrated n Table I. Fgure 1. Telecontrol loop along wth control staton and robot. (2) k When the rule s represented by Mamdan (conjuncton relaton) and snce our case has two antecedents (The Joystck sgnal and Dstance from sonar sensors) then, the Mamdan nference s obtaned n (3). A (u ) A (u )... c* ( w) B (v) B (v)... u v C ( w) w W Fgure 2. Sonar sensors mounted n front of the robot. TABLE I. Sonar Readngs (3) VC C M F VF In the Defuzzfcaton stage that converts the fuzzy set obtaned from nference mechansm process nto crsp value by Center of Gravty as descrbed n (4). FUY RULES BASE FOR SPEED CONTROLLER 19 Joystck Command HN LN LP HP PH

4 Internatonal Journal of Mechancal Engneerng and Robotcs Research Vol. 5, No. 1, January 2016 Fg. 11 shows the performance of controller, the robot moves accordng to the commands from the human operator but once t detects the obstacle t reduces ts speed regardless the value of nput from human operator, and then stops when t goes very close to the obstacle and only follow the human operator command n backward drecton and gnore the command for forward moton. Fg. 12 and Fg. 13 shows the teleoperaton usng wreless network n the RST Insttute. The robot s calculatng the output speed based on the envronment. Fgure 3. Fuzzy membershp functons for nput1. VI. CONCLUSION AND FUTURE WORK In ths paper, we have presented fuzzy control technque to cope wth the ssue of delay tme from the Internet communcaton for ground robot teleoperaton. The expermental results showed that the controller acheved the requrements of the project and t ensured the stablty and relablty to accomplsh the actual exploraton mssons. In addton, the results showed excellent expermental performance of the ground robot navgaton. For the future work, we would lke to nculcate servo control to steer the robot n desred left or rght drecton. Vson based navgaton s also under plannng, whch wll be used for trackng and detecton purposes. Fgure 4. Fuzzy membershp functons for nput2. Fgure 5. Fuzzy membershp functons for output. Fg. 6 demonstrate the fuzzy surface generated by two nputs and one output. The surface shows the output speed based on nput from joystck and sonar readng. V. Fgure 6. Fuzzy surface. EXPERIMENTAL RESULTS Several experments were carred out to analyze the behavors of the controller. The robot was tested n ndoor envronment and results have been plotted to show the performance durng the experments. Fg. 7, Fg. 8, and Fg. 9, are llustratng the values of sensors readngs. Fg. 10, s showng fuzzy speed controller output speed. Blue lne plot s representng the sensor S1, red lne plot shows sensor S2, dark green plot shows sensor S3 and lght green plot s showng the output speed. When there s no obstacle then the sensors are showng the maxmum dstance.e. 200cm. In ths case the robot s followng the human operator and the varaton n speed s beng brought by the operator command accordng to the fuzzy rules defned n the Table I Operator s drvng the robot n forward or backward drectons but when the objects are beng detected by the sensors n the near vcnty of robot then the speed s reduced as shown by lght green plot n Fg. 10, and the robot stops when t goes very close to obstacle as shown n lght green plot after 120 samples. Fgure 7. Readngs of sensor 1. Fgure 8. Readngs of sensor 2. 20

5 Internatonal Journal of Mechancal Engneerng and Robotcs Research Vol. 5, No. 1, January 2016 REFERENCES [1] [2] Fgure 9. Readngs of sensor 3. [3] [4] [5] [6] Fgure 10. Output speed. [7] [8] [9] [10] Fgure 11. Robot detectng human obstacle and reducng speed. [11] [12] [13] [14] Fgure 12. Robot navgaton n lab. [15] [16] [17] [18] [19] [20] Fgure 13. Robot navgatng n ndoor envronment. 21 P. Falcon, A. Barrero, M. D. Cacho, and E. Delgado, Passve teleoperaton of moble robot wth nput-output lnearzaton and dynamc extenson, n Proc. 4th Annual IEEE Internatonal Conference on Cyber Technology n Automaton, Control and Intellgent Systems, Hong Kong, Chna, 2014, pp I. Farkhatdnov and J. Ryu, Improvng moble robot blateral teleoperaton by ntroducng varable force feedback gan, n Proc. IEEE/RSJ Internatonal Conference on Intellgent Robots and Systems, Tape, Tawan, 2010, pp O. Lnda and M. Manc, Self-organzng fuzzy haptc teleoperaton of moble robot usng sparse sonar data, IEEE Transactons on Industral Electroncs, vol. 58, no. 8, I. Farkhatdnov, J. Ryu, and J. An, A prelmnary expermental study on haptc teleoperaton of moble robot wth varable force feedback gan, n Proc. IEEE Haptcs Symposum, Waltham, Massachusetts, USA, 2010, pp L. hang, C.. Xn, and W. Ja, A networked teleoperaton system for moble robot wth wreless seral communcaton, n Proc. IEEE Internatonal Conference on Intellgent Computng and Intellgent Systems, Shangha, Chna, 2009, pp A. H. Dallal, A. S. Derbala, and M. F. Taher, A moble robot for hosptals controlled usng wlan, n Proc. Internatonal Bomedcal Engneerng Conference, Caro, Egypt, 2012, pp M. Kasm, Al-Aubdy, M. M. Al, A. M. Derbas, and A. W. AlMutar, Gprs-based remote sensng and teleoperaton of a moble robot, n Proc. 10th Internatonal Mult-Conference on Systems, Sgnals & Devces, Hammamet, Tunsa, J. Babc, M. Buds, and I. Petrovc, Dynamc wndow based force reflecton for safe teleoperaton of a moble robot va Internet, n Proc. IEEE/ASME Internatonal Conference on Advanced Intellgent Mechatroncs, urch, 2007, pp Y. Wagatsuma, Y. Toda, and N. Kubota, Formaton behavor of multple robots based on tele-operaton, n Proc. IEEE Internatonal Conference on Fuzzy Systems, Tape, Tawan, 2011, pp N. C. Mtsou, S. V. Velanas, and C S. Tzafestas, Vsuo-haptc nterface for teleoperaton of moble robot exploraton tasks, n Proc. 15th IEEE Internatonal Symposum on Robot and Human Interactve Communcaton, Hatfeld, UK, 2006, pp S. A. Kayan, W. H. Bhatt, and K. K. Jarral, On desgn and fabrcaton of a prototype teleoperated moble survellance robot, n Proc. IEEE Internatonal Conference on Emergng Technologes, Islamabad, Pakstan, 2009, pp S. K. Cho, H.. Jn, J. M. Lee, and B. Yao, Teleoperaton of a moble robot usng a force-reflecton joystck wth sensng mechansm of rotatng magnetc feld, IEEE/ASME Transactons on Mechatroncs, vol. 15, no. 1, pp , N. Dolat and C. Melchorr, Tele-operaton of a moble robot through haptc feedback, n Proc. IEEE Internatonal Workshop on Haptc Vrtual Envronments and Ther Applcatons, Ottawa, Ontaro, Canada, Y. Kawa, T. Namerkawa, and M. Fujta, Blateral teleoperaton of wheeled moble robot wth tme delay usng vrtual mage robot, n Proc. IEEE Internatonal Conference on Control Applcatons, Yokohama, Japan, M. Boukhnfer, A. Chabet, T. Azb, and C. Larouc, Tele-control electrcal vehcle system usng wave varables, n Proc. 9th Internatonal Mult-Conference on Systems, Sgnals and Devces, Chemntz, J. S. R. Jang, C. T. Sun, and E. Mzutan, Nuero Fuzzy and Soft Computng, Upper Saddle Rver, NJ: Prentce-Hall, 1997, pp L. adah, Fuzzy sets, Informaton and Control Journal, vol. 8, pp , L. A. adeh, The concept of lngustc varable and ts applcaton to approxmate reasonng-i, Informaton Scences, vol. 8, pp , L. A. adeh, The concept of lngustc varable and ts applcaton to approxmate reasonng-ii, Informaton Scences, vol. 8, pp , K. Demrl and I. B. Turksen, A revew of mplcaton and the generalzed modus ponens, n Proc. 3rd IEEE Conference on

6 Internatonal Journal of Mechancal Engneerng and Robotcs Research Vol. 5, No. 1, January 2016 Fuzzy Systems, IEEE World Congress on Coputatonal Intellgence, 1994, pp Aamr Shahzad was born n Gujranwala Pakstan. He has receved hs BSc Mechancal Engneerng degree n 2006, from Unversty of Engneerng and Technology Lahore, Lahore, Pakstan. Then he earned the degree of master n Mechatroncs n 2011 from Unversty of Segen, Segen, Germany. He has worked n RCET Gujranwala, a consttuent college of UET Lahore as Lecturer for two year from ( ). And currently a PhD student and research assstant at Insttute of Automatc Control Engneerng (RST), Unversty of Segen, Segen, Germany. Hs research nterests are moble robot telecontrol, teleoperaton, embedded systems and network programmng et al. Segen, Segen, Germany He worked as a lecturer n Mechancal Engneerng Department at Jordan Unversty of Scence & Technology and n Techncal & Vocatonal Tranng Corporaton, Computer Engneerng Department, Saud Araba, Hubert Roth was born Germany, He completed a Doctorate degree at Karlsruhe Techncal Unversty, Germany n Currently he s the Char n Control Engneerng Department (RST) at Segen Unversty, Faculty of Electrcal Engneerng and Informatcs snce Prof. Dr.-Ing. Roth s a vce Drector of Techncal Commttee on Computers and Telematcs of Internatonal Federaton of Automatc Control (IFAC) snce 2008 and vce Drector of Centre for Internatonal Capacty Development (CICD) snce Ram Al-Jarrah was born n Jordan. He completed a PhD degree n Robotcs and Intellgent Control at Automatc Control Engneerng Insttute (RST), Segen Unversty, Germany n Hs research nterests are Fuzzy Logc analyss, Image Processng and Vson System, Fuzzy Image Processng, Embedded System, Wreless Sensor Networks and Aeral Robotcs. Dr.-Ing. Al-Jarrah worked a research assstant at Insttute of Automatc Control Engneerng (RST), Unversty of 22