2016 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media,

Transcription

1 2016 IEEE. Personal use of ths materal s permtted. Permsson from IEEE must be obtaned for all other uses, n any current or future meda, ncludng reprntng/republshng ths materal for advertsng or promotonal purposes, creatng new collectve works, for resale or redstrbuton to servers or lsts, or reuse of any copyrghted component of ths work n other works.

2 A Telepresence Wheelchar Usng Cellular Network Infrastructure n Outdoor Envronments Van Kha Ly Ha, Student Member, EMBS, IEEE, Tuan Ngha Nguyen, Senor Member, IEEE, and Hung T. Nguyen, Senor Member, IEEE Abstract Moble wreless network technology has grown rapdly over the past decade wth emergng applcatons and servces. Partcularly, the fourth generaton (4G) cellular network has acted as a brdge between telecommuncaton technology and daly lfe applcatons. In ths paper, we present an nvestgaton nto a telepresence wheelchar n outdoor envronments employng cellular network nfrastructure nstead of usng local wreless networks n ndoor envronments. Experments were carred out to demonstrate remote nteracton and control from a long dstance and across countres. A large amount of communcaton data based on real network measurements was collected and analyzed to evaluate the system performance. The expermental results show that a wheelchar system can be controlled remotely n real-tme wth the acceptable round trp tme of less than 400 ms. The results reveal the feasblty of usng the 4G network for a telepresence wheelchar n healthcare applcatons. I. INTRODUCTION In the modern world, nnovatons n telecommuncatons have made sgnfcant mpacts on human modern lfe and offered numerous applcatons n daly lfe. However, the use of advanced telecommuncaton technologes to develop solutons for mprovng healthcare such as tele-montorng, socal nteracton, and remote healthcare assstance through a wheelchar for people wth dsabltes s lmted due to ts techncal challenges [1]. Recently, smart wheelchars for people wth dsabltes have attracted consderable attenton. There are a wde varety of smart wheelchars that have been developed by researchers [2], and references theren. The focus of these studes has been on drect assstance whle a wheelchar user s sttng n a wheelchar. Our work focused on the control of wheelchars usng head movement, bran waves, and sensors. One typcal example [3], descrbed an ntellgent wheelchar whch works as a hands-free wheelchar control system and provdes obstacle detecton whle navgatng through an unknown envronment to assst wheelchar users. However, the ablty to control the moblty of wheelchars from a remote locaton, long dstance, has not been studed n the lterature thus far, although such meanngful systems are useful for people wth dsabltes. Van Kha Ly Ha, Tuan Ngha Nguyen, Hung T. Nguyen are wth the Centre for Health Technologes, Faculty of Engneerng and Informaton Technology, Unversty of Technology Sydney, NSW 2007, Australa, (emal: VanKha.L.Ha@student.uts.edu.au; tuanngha.nguyen@uts.edu.au; Hung.Nguyen@uts.edu.au). Much research has been done n the feld of telepresence. Telepresence nvolves many technologes such as robotcs, computer and telecommuncaton networks. Apart from telepresence robots for meetngs, telepresence robots for educaton, and for vstng people have been explored n [4] and [5], respectvely. These telepresence robots were developed through the addton of a webcam, computer, and moble robot platform [4, 5]. To date, current telepresence works have been lmted n the feld of robotc applcatons and beng used only n ndoor envronments. The advancement of telepresence technology has not stopped, though. People wth dsabltes would beneft from the use of these technologes. There s a gap n the lterature wth regard to developng a so-called telepresence wheelchar. The use of telepresence wheelchars could provde moblty assstance to people wth dsabltes when dstance separates the partcpants. Ths system wth realtme vdeo nteracton would allow a remote user to nteract, and assst a dstant wheelchar user. The system not only provdes healthcare applcatons to wheelchar users, but also acts as a brdge between them and remote users, doctors, and ther relatves. The man objectve of ths paper s to present our practcal nvestgaton of a telepresence wheelchar usng cellular networks. Our approach mplctly explots technologcal advancement n telecommuncaton, automaton control, and healthcare systems. Our key contrbutons demonstrate the feasblty of a telepresence wheelchar for face-to-face communcaton and effcent remote control va a cellular network medum. In partcular, we have developed approprate software for two-way vdeo nteracton and telepresence wheelchar remote control based on the Skype framework. We conducted real-tme experments n a long dstance connecton wth the functon of telepresence n outdoor envronments. The remote user can drve the wheelchar va the Internet. At the same tme, the remote user can hear sound and see the surroundng areas of the wheelchar wth the 360-degree vew n real-tme. The performance of such an nteractve telepresence wheelchar usng a 4G moble network was consdered. Techncal performance measurements were conducted to collect transmsson data for evaluaton of the system performance. The remander of ths paper s organzed as follows. Secton II descrbes the desgn of the telepresence wheelchar system, the Skype framework, and the 4G cellular network. Secton III s devoted to the experments, results, and dscusson. Fnally, conclusons and future works are drawn n Secton IV.

3 Mn Computer Touch Screen Dsplay II. METHODOLOGY A. Desgn of the Telepresence Wheelchar The telepresence wheelchar s a system that provdes realtme audo and vdeo communcaton capabltes, and can be controlled remotely from a long dstance. The telepresence wheelchar platform has been developed as shown n Fg. 1. In order to acheve the desred features, the powered wheelchar s equpped wth a Mac mn computer, control components, telecommuncaton nterfaces, touchscreen dsplay and four webcams for a wde vew surroundng the wheelchar. Partcularly, a NI USB-6008 s connected on the back of the wheelchar to transform movement commands from the remote user to control the wheelchar. NI USB s controlled by an applcaton that can be plugged nto Skype framework. 4G Modem Fgure 1. The back and front vew of the telepresence wheelchar. B. Skype Framework NI USB-6008 Multple Webcams Skype has changed the world wth ts free voce-over-ip servce, provng hgh-qualty two-way vdeo communcaton over the Internet [6]. There are other frameworks for telepresence such as the Server and Clent webste, and WebRTC. However, Skype s a well-known cross-platform, snce users can expand usng an Applcaton Programmng Interface (API). It s an deal framework for constructng an nexpensve system. For that reason, we have developed a telepresence wheelchar based on ths ubqutous framework. Our approach s adapted to the specfc applcaton n the feld of health care. To be able to acheve the remote nteracton characterstcs, a control nterface wth fve buttons, namely go forward, turn left, turn rght, go backward and stop was desgned to send command sgnals to Skype. These control sgnals are wrtten and read by an API developed from C#. At the wheelchar sde, these commands wll be receved and transmtted to the NI USB 6008 to drve the wheels n realtme. An applcaton wth the functons of acqurng, dsplayng a lve vdeo stream and controllng the wheelchar was developed. In ths applcaton, once the connecton s establshed, a remote user wll be able to vew lve vdeo around the wheelchar and communcate wth a wheelchar user by usng the 4G cellular network. C. 4G moble network archtecture The development of the cellular network has revolutonzed the transfer of nformaton. Partcularly, the fourth generaton cellular network (4G) or Long-Term Evoluton (LTE). The LTE s a standard for wreless communcaton of hgh-speed data for moble phones and data termnals developed by the 3rd Generaton Partnershp Project (3GPP) [7]. In theory, 4G ncreases the capacty and speed usng a dfferent rado nterface together wth core network mprovements wth the peak of download speeds up to 100Mbps and 50Mbps upload. The LTE network archtecture dagram s shown n Fg. 2. It llustrates the connecton paths from the end user equpment (UE) to the evolved base statons (enodeb). Each enodeb s a base staton that communcates wth and controls the mobles. The moblty management s responsble for choosng the servng gateway (S-GW). S-GW acts as a router, and exchanges data between the base staton and the packet data network gateway (P-GW), and communcates wth the outsde world such as the Internet or the IP networks. The telepresence wheelchar requres good qualty vdeo streamng and stable communcaton lnks between geographcally dstrbuted users. Understandng the network performance s mportant for an effectve telepresence system. The control processng capabltes of the wheelchar depend on the transmsson delays and other factors. In partcular, due to moblty ssues and rado condtons, fluctuaton sgnals and the lack of bandwdth cause latency and jtter whch can mpact the overall system performance. Latency or round-trp tme (RTT) refers to the amount of tme t takes a bt from transmsson to recepton. The large RTTs result n system connecton tme-out. Jtter causes packets to arrve at ther destnaton wth dfferent tmng. RTT and jtter calculaton has been presented n [8]. It denotes the correspondng event tmes as Ts and Tr for sendng and recevng tme respectvely (for packet ID = ). The RTT s the dfference between the event tmes RTT Tr Ts (1) For jtter calculaton, expresson (2) s used. In ths formula the ndvdual latency measurement samples for two subsequently receved packets of the same stream are defned as RTT, RTT and the dfference of these two latency j measurement results s defned as J j, : 1 2 J 0, J 2,1 J 1 3 J 1 J, 1 J 16 The factor of 16 n the denomnator of (2) counts for a smoothng factor proposed n the standard RFC 1889 [8]. In ths study, a moble broadband data 4G modem s plugged n the wheelchar and acts as a UE for telecommuncaton nterfaces as shown n Fg. 1. Therefore, the telepresence wheelchar can establsh connectons to the Internet va the telecommuncaton nfrastructures for transferrng vdeo and control sgnals. (2)

4 LTE Rado Acess Network enodeb 3G Network UE Moblty Management S-GW LTE P-GW Internet / IP network Wheelchar enodeb Remote user Fgure 2. A smplfed LTE cellular network archtecture. A. Expermental setup III. EXPERIMENTS AND RESULTS The expermental study on the performance of the telepresence wheelchar was deployed n outdoor envronments. The wheelchar was located wthn the pedestran walkway at Jones Street n Sydney, Australa. The system confguratons of a remote user and wheelchar are lsted n Table I. The wheelchar computer was confgured wth a 4G modem on the Telstra moble network n Sydney, New South Wales, Australa. In ths setup, a wheelchar user was requred to st n the wheelchar wthout dong anythng and remote users connected, nteracted and navgated the wheelchar as shown n Fg. 3a. We conducted two case studes. Frst, we carred out experments wth domestc connectons n the same country. In ths scenaro, remote users and a wheelchar user would be n Australa. In the second scenaro, we set up other experments to connect and control the wheelchar wth the functon of telepresence from a dfferent country n a long dstance connecton. The remote user was n another country (Vetnam) and the wheelchar user was n Australa as shown n Fgs. 3a, 3b and 4. For our system performance to be analysed and evaluated, several measurements were taken at each ste durng the connecton to measure the mpact factors for ongong connectons. Four key performance parameters were taken nto account, ncludng sgnal strength, round-trp tme, jtter and central processor unt usage. The results were used to analyse the performance of our system and made comparsons wth prevous studes. Wreshark was used to capture and analyse network traffc [9]. Pngs were smultaneously sent from the remote user to the wheelchar for RTT measurements, and the respectve latency data was saved nto a log fle. TABLE I. DETAILS OF THE COMPUTERS USED FOR IMPLEMENTATION. System nformaton Remote computer Wheelchar computer Model MacBook pro Mac mn CPU, DRAM Intel Core 7, 8 GB Intel Core 7, 8 GB OS Wndow 7 (64 bt) Wndow 7 (64 bt) Number of webcams 1 4 Skype Verson Verson Wreless W-F 4 G Transport protocol TCP/IP, UDP TCP/IP, UDP Movement speed 0 1m/s (a) Fgure 3. (a) Wheelchar n outdoor envronment at Jones street Sydney. (b) A connecton from remote user n Vetnam to wheelchar user n Sydney Fgure 4. Remote user controlled wheelchar from a long dstance. B. Results and Dscusson Rado recepton strength of the 4G modem on the computer of the wheelchar was measured n reference sgnal receved power (RSRP) n decbel mllwatt (dbm) at a dfferent tme whle the wheelchar was movng. The measurement results are descrbed n Fg. 5. RSRP vared between -93 and -101dBm for the frst scenaro and -93 and 107dBm for the second scenaro. In comparson wth [10], the moble recepton sgnal of the wheelchar was suffcent for a remote user to connect to the wheelchar and control the wheelchar wth telepresence functons. Fgure 5. The RSRP measurement from the wheelchar of experments Fgure 6. Round trp tme vs jtter over a perod of 4500 mllseconds of same country connectons (Australa Australa). (b)

5 Fgure 7. Round trp tme vs jtter over a perod of 4500 mllseconds of dfferent country connectons (Vetnam Australa). Fgure 8. CPU usage of wheelchar computer and remote user s computer. Through a seres of careful measurements, the average RTT and jtter measurement results for local and global connectons are shown n Fg. 6 and 7. It can be observed that round trp tmes fluctuated consderably between 226 and 271 mllseconds (ms) for local connectons. RTTs of global connectons were more than twce as hgh as that of local connectons. In some cases, the global connecton RTTs were n excess of 400 ms. It s worth notng that the round trp tme of less than 400 ms s consdered as real-tme vdeoconference [11]. In comparson wth a smlar prevous study [12], our proposed system has superor performance. Regardng the jtter of real-tme traffc, there are sgnfcant dfferences between the local and nternatonal connectons. The mnmum jtter n the local connecton s 20 ms and the mnmum jtter n the nternatonal connecton s under 60 ms. Moreover, durng our experments, we also found that the performance of central processor unt (CPU) of the wheelchar computer and remote computer made a bg dfferent n terms of CPU usage. Ultmately, all these parameters are correlated to the mpact of moblty ssues n system performance. There s a correlaton also between the dstance, rado recepton strength, and delay. The data shows greater latency n long dstance connectons than n local connectons. Whle these fndngs may not come as a surprse to most, t can be enhanced by mplementng dversty network optmzatons n the 4G networks. However, ths study s RTT s stll under the theoretcal RTT threshold value [11]. It ndcates that the latency mpact of the cellular network s acceptable for remote users to control the wheelchar. Our work demonstrates the successful expermental deployment of a telepresence wheelchar n an outdoor envronment assocated wth a certan scenaro. It was observed that n some cases the cellular network lnks dsconnected and reconnected due to sgnal loss, rado nterference, and cell re-selecton. Nevertheless, the expermental results obtaned have demonstrated the potental of a telepresence wheelchar usng 4G cellular networks. Our future research wll conduct experments n the weak sgnal envronments. The status of the wheelchar, the state of the wheelchar battery, the feedback nformaton, the poston trackng, and the user s bology sgnal would also be ntegrated. Ths system not only provdes healthcare applcatons to wheelchar users but also acts as a brdge between wheelchar users, remote doctors, and ther relatves. These postve outcomes may help the robotc and wheelchar communty to make a sgnfcant nvestment n developng the telepresence wheelchar for healthcare. IV. CONCLUSION We have presented the feasblty of the deployment of the telepresence wheelchar system n outdoor envronments. The expermental results obtaned on the wheelchar n a practcal envronment are postve and relable, whch shows potental for telepresence nnovatons n the health care doman. The results demonstrated the effectveness of the combnaton of advanced technologes for a telepresence wheelchar. The system allows a remote user to control the wheelchar remotely and communcate wth clents from a dstant locaton through voce and vdeo. Future work wll consder transmttng feedback nformaton and bo-sgnals to enable real-tme health montorng. REFERENCES [1] V. K. L. Ha, T. N. Nguyen, and H. T. Nguyen, "Real-tme transmsson of panoramc mages for a telepresence wheelchar," n 37th Annual Internatonal Conference of the IEEE Engneerng n Medcne and Bology Socety (EMBC), 2015, pp [2] A. V. Nguyen, L. B. Nguyen, S. Su, and H. T. Nguyen, "The advancement of an obstacle avodance bayesan neural network for an ntellgent wheelchar," n 35th Annual Internatonal Conference of the IEEE Engneerng n Medcne and Bology Socety (EMBC), 2013, pp [3] J. S. Nguyen, N. Tuan Ngha, Y. Tran, S. W. Su, A. Crag, and H. T. Nguyen, "Real-tme performance of a hands-free sem-autonomous wheelchar system usng a combnaton of stereoscopc and sphercal vson," n Annual Internatonal Conference of the IEEE Engneerng n Medcne and Bology Socety (EMBC), 2012, pp [4] S. C. Herrng, "Telepresence Robot for Academcs," presented at the ASIST, Montreal, Quebec, Canada, [5] K. Tatsuno, T. Kawa, M. Nako, Y. Yasuda, T. Fukuta, and H. Murata, "Development of a remote vstor robot system" n Annual Conference 2010, Proceedngs of SICE, 2010, pp [6] Mcrosoft Skype. (2015, 10 Oct). Avalable: [7] G. Swach and A. Esmalpour, "LTE Securty potental vulnerablty and algorthm enhancements," n 7th Canadan Conference on IEEE Electrcal and Computer Engneerng (CCECE), 2014, pp [8] R. Kreher and K. Gaenger, LTE Sgnalng, Troubleshootng, and Optmzaton, John Wley & Sons, Ltd, [9] A. Orebaugh, G. Ramrez, J. Burke, and L. Pesce, Wreshark \& Ethereal Network Protocol Analyzer Toolkt (Jay Beale's Open Source Securty): Syngress Publshng, [10] P. T. P. Ltd. (2014, 30 Jan). Moble network gude mprovng moble sgnal. Avalable: [11] ITU-T. Recommandaton. G.1010, "Qualty of servce and performance", n Transmsson systems and meda, dgtal systems and networks, Geneva, Swtzerland, [12] J. D. Terrazas Gonzalez, G. Lnton, F. Wa-keung, and R. Barwell, "On nternet based supermeda enhanced telepresence va cellular data network," n 38th Annual Conference on IEEE Industral Electroncs Socety IECON, 2012, pp