Archive of Photogrammetry, Cartography and Remote Sening, Vol., 0, pp. 39-49 ISSN 083-4 ASSISTING PERSONAL POSITIONING IN INDOOR ENVIRONMENTS USING MAP MATCHING Mohamed Attia, Adel Moua, ing Zhao 3, Naer El-Sheimy 4,, 3, 4 Mobile Multi-Senor Sytem MMSS Reearch Group, Department of Geomatic Engineering, Univerity of Calgary, Alberta, Canada, TNN4 maaattia, amelaye, xinzhao, elheimy@ucalgary.ca KE WORDS: Multi enor, GPS/INS, GIS, Real-time, Matching, Navigation ABSTRACT: Peronal poitioning i facing a huge challenge to maintain a reliable accuracy through all application. Although in outdoor application, everal mobile navigation device can provide acceptable poitioning accuracy, the ituation in indoor environment i not the ame. Mobile navigation device mainly contain a global poitioning ytem GPS receiver and an inertial meaurement unit IMU. The main drawback in indoor navigation application i the unavailability of the GNSS ignal, which decreae the poibility of obtaining an accurate abolute poition olution, a the inertial ytem INS olution will drift with time in the abence of external update. Several alternative were preented lately to update the inertial olution uch a uing Wi-Fi, UWB, RFID, everal elf-contained enor, imaging aiding and patial information aiding. In order to achieve accurate poition olution, with low-cot and uable technique, an integrated mobile navigation ytem integrating GPS/IMU/Wi-Fi and map-matching wa developed. The developed ytem ue the prior knowledge of the indoor geometrical and topological information, a a threhold for the navigation olution, forcing the provided olution to be motly on the right track. The geometrical and topological information for the building wa ued to build the geopatial data model. The ue of thi model wa performed by developing a map matching algorithm which ue the geometrical and topological characteritic of the building to locate the uer poition on the building map. Thi algorithm wa developed baed on the geopatial information of the Engineering building, Univerity of Calgary, where the field tet occurred. The map-matching algorithm wa evaluated by proceing and comparing two eparate navigation olution through the tudy area, one uing only the GPS/IMU/Wi-Fi ytem, and econd olution wa aited with the map-matching algorithm which how ignificant enhancement in the poition olution for the indoor trajectory.. INTRODUCTION Navigation ytem play an important role in many vital dicipline. Determining the accurate location of a uer relative to the phyical environment e.g. roadway, interection, ervice i an important part of a range of tranportation ervice, for example in-vehicle navigation, fleet management and infratructure maintenance Syed, 009. Alo other ervice require locating a uer to many phyical indoor environment e.g. airport, hopping mall, public building to ait in many navigation application uch a E9, law-enforcement, handicapped movement and marketing ervice. Since all thee application require a truted and reliable poitioning, whenever the uer i located in a GPS denied environment, there i a huge challenge to maintain thi reliability. Several reearche addre poible methodologie to aid the current navigation ytem in indoor 39
Mohamed Attia, Adel Moua, ing Zhao, Naer El-Sheimy environment with either additional enor or pecial filtering, however reaching an abolute accurate and truted poition in thee environment i a challenge till thi moment El-Sheimy, 00. The emerging current technologie uitable for indoor navigation include ground-baed RF ytem, uch a UWB and RFID; maintream wirele communication ytem, uch a cellular phone and WLAN e.g. Wi-Fi; viion aiding ytem, uch a camera and laer; and dead reckoning enor, uch a gyrocope and accelerometer Zhao, 00. Conidering the availabilitie, accuracy and cot factor for ma production, the above candidate technologie face ome limitation for practical implementation. Therefore it i a mut to depend on multiple technologie a well a earch for the optimal integration between thee technologie, depending on the application, and ait them by other mean if applicable El-Sheimy, 00. Mot of the navigation device ue a map matching technique to preent the uer location on the map. Map matching itelf ha been developed to be more than jut a viualization tool to locate the navigation data poition and direction on a map. Mitigating the error from navigation enor and bridging the navigation olution have become one of the main tak for any map matching algorithm Attia, 00. In indoor environment, where it i hard to provide a continuou reliable poition etimate, map matching can bridge the navigation olution baed on the floor alignment. The building information provide a logical threhold to bound the olution into a certain region, changing the main target of the navigation ytem from obtaining a high accuracy to poition information to obtaining a poition with enough accuracy allow the ytem to elect the correct paageway. Thi paper will addre aiding indoor navigation ytem with mapping information, getting benefit from the geopatial information for the navigated region, i.e. floor plan. The navigation ytem ue a real-time hardware peronal navigation ytem PNS developed at the Univerity of Calgary. The PNS unit integrate GPS and microelectronic gyrocope/accelerometer along with Wi-Fi Both performance and cot of thi unit i comparable to the enor ued in mart cell phone Zhao, 00. Within the indoor environment, the GPS doe not have any ignificant benefit through the navigation olution other than to initialize the poition when the uer enter the building. Peronal Dead Reckoning PDR wa ued to integrate the enor data with the Wi-Fi data. A map matching algorithm wa developed to locate the poition fix from the navigation olution on a developed geopatial data model for the navigated building. The algorithm ue geometrical and topological contraint to project the poition on the nearet paageway taking into conideration the connectivity of the paageway. The main objective of thi integrated algorithm i to provide a more reliable navigation olution working in indoor environment. A brief background on related work, the decription of the ytem, teting environment and reult will be dicued through the following ection.. MAP AIDING NAVIGATION SSTEMS Map have been ued for centurie to tranit uer from a certain place to another. For the lat decade, navigation device have ued the digital form of map to locate the poition of the uer on them in order to ait in providing the navigation direction. Recently map have become more than jut a viualization tool in navigation ytem, but they have been an aiding tool in enhancing the reliability of the obtained navigation olution. Navigation 40
Aiting peronal poitioning in indoor environment uing map matching ha a huge et of application, which motly ue map in their diplay, uch a in-vehicle application, peronal navigation device, and even mart phone with navigation application. Thee application uually ue a navigation ytem coniting of a GPS receiver, a map and it geopatial databae. Although, thoe ytem uually provide accurate poitioning information, it i limited to the open pace environment. Navigation in GPS denied environment uch a indoor facilitie are facing a huge challenge to maintain the accurate poitioning information due to the GPS ignal blockage and multipath Syed, 009. Aiding thee ytem by integrating other navigation enor uch a INS i ueful in ome cae, but the drift in poition error with time due to GPS outage, till reduce the reliability of thi olution in ome application. The reearch of aiding navigation ytem uing map matching deal with three apect; electing the integrated navigation enor, deigning the map matching algorithm and deigning the geopatial model for the floor plan. For indoor application and due to the ignal blockage, GPS/INS integrated ytem require another aiding ource a dicued in the introduction ection. For the econd apect, Quddu, 007 compare between everal poible map matching algorithm according to their accuracy and concept. Generally, there are three type for map matching; geometrical and topological algorithm, probability algorithm and advanced algorithm uch a uing fuzzy logic, belief theory and Bayeian network White, 000; Greenfeld. 00; Quddu, 003, Banayake, 005. The main difference between thee categorie i the method in electing the matching link where the poition will be projected on. The third apect i chooing the appropriate map deign which ha the characteritic to upport navigation application Bullock, 994. Map for indoor application hould model all poible paageway and height change acce tair, elevator, and alo conider the topological characteritic, for example connectivity between paageway. Mieenberger, 008; Arto, 009 dicu the requirement for deigning navigation baed map. Gillieron, 003, Khider, 008; Glanzer, 009 dicu ome ytem for indoor navigation with the ait of the building information. The main idea i to ue the building information to create virtual boundarie for the navigation olution, which lead to bounding the olution in the mot probable region, therefore increaing the reliability of the olution. 3. METHOD A tated previouly, the propoed algorithm baed on three main component; the navigation ytem, the geopatial data model and the map matching algorithm. The integration between thee three component i illutrated in figure. For the navigation ytem, the tet ytem wa et up coniting of a laptop with WLAN Mini-card, Garmin CS60 GPS and ADI ADIS605 IMU. The IMU contain tri-axial digital gyrocope and accelerometer. All the enor and GPS data are collected from USB port and the computer ynchronize all the data internally. The Wi-Fi poitioning algorithm work on updating the navigation olution when GPS i not available, which i the cae in mot indoor environment. Wi-Fi poitioning i baed on the meaurement of received ignal trength RSS from the acce point AP, by uing ignal propagation model to convert ignal trength to a ditance meaurement to the acce point. 4
Mohamed Attia, Adel Moua, ing Zhao, Naer El-Sheimy A looely-coupled Extended Kalman Filter i applied to integrate tri-axial accelerometer and gyrocope with Wi-Fi and GPS update. The algorithm tart by uing both accelerometer and gyrocope data which i integrated with GPS meaurement once they become available. During good GPS/Wi-Fi update, mialignment attitude angle between body frame and peron frame are etimated El-Sheimy, 00. A Pedetrian Dead Reckoning PDR i adopted to minimize enor integration error and drift by exploiting the kinematic of human gait with the traveled ditance and heading information Zhao, 00. The algorithm contain tep detection, tride length etimation and heading determination. Once mechanization derived heading i converged, the heading i ued in PDR to compute Eating and Northing for the Kalman Filter; however if heading drift too much, or if tep aren t detected with enough certainty, only mechanization i ued. More detail on the navigation ytem are dicued in Zhao, 00. 4 Fig.. Propoed map aiding algorithm A geopatial data model wa developed for the tudy area, which i the engineering building at Univerity of Calgary. The floor map, provided from the Univerity geographical information library, wa filtered to repreent the region urrounding the field tet area. The geopatial model include the geometrical modelling for the paageway, and the aociated attribute. The paageway corridor were modelled a polyline link with two node; the tart node and the detination node. All node were decribed by an identification number ID and a projected UTM coordinate. The aociated attribute for the link include the floor number, proximity of a height change acce, tair, and all poible link diverged from it tart and detination node. Figure how the geometrical layout for the model.
Aiting peronal poitioning in indoor environment uing map matching Fig.. Geopatial model for the tudy area Finally, a map matching algorithm wa deigned to locate the poition fix from the navigation ytem to the map. The map matching algorithm i baed on the geometrical and topological algorithm developed in Attia, 00 with ome modification to fit the indoor environment. Two different matching logic were developed. Both map matching algorithm are baed on point to curve matching Quddu, 007. The point to curve map matching i baed on projecting the poition etimated from the navigation algorithm S, S into the nearet link, which in thi cae i the nearet paageway. Each paageway link ha a tart and end node, [,,, ]. Uing the poition etimate coordinate S, S a dot product between the poition and the tart and detination node [,,, ] i done a in equation to calculate the minimum ditance between the poition etimate and all the link provided that the projection lie within the extent of the link. Equation i baed on the concept that the minimum ditance between any point and a line i the perpendicular ditance which i obtained uing the dot product. After applying equation on all the link, and once a link i elected, the algorithm will project the poition etimate on the paageway link. The projected coordinate P, P can be obtained uing equation, 3 Quddu, 003. 43
44 Equation and 3 imply calculate the coordinate in and repectively for projection of the poition etimate S, S on the elected nearet paageway whoe tart and end node are [,,, ]. D = ] [ p = ] [ p = 3 The difference between the two algorithm i how the navigation olution i projected. A for the firt algorithm, it will read the firt epoch poition and project it on the nearet link uing the previou technique. It will then keep reading epoch by epoch and project them on the nearet link. The econd algorithm will tart imilarly to the previou one; however it will compute the hift between the firt poition etimate and the projected poition. The algorithm will hift the whole navigation olution with thi value before projecting the econd epoch. Therefore for every epoch, the algorithm will firt hift the olution with the previou epoch hift value, and then project it poition on the nearet link. Figure 3 preent the outline and difference between the two algorithm. The reult for both algorithm are illutrated in the next ection. Fig. 3. The outline for firt algorithm on right and the econd algorithm on left Mohamed Attia, Adel Moua, ing Zhao, Naer El-Sheimy
Aiting peronal poitioning in indoor environment uing map matching 4. FIELD TEST AND RESULTS Pedetrian field tet were performed in a number of indoor and outdoor cenario to further verify the performance of thi algorithm. The ued dataet wa collected in January 00 at engineering building, Univerity of Calgary campu. It i a ten minute walk from outdoor to indoor, a hown in figure 3. The tet tarted at the eat door entrance of the building; the uer then walked two loop outdoor and entered the building; then the uer walked indoor and went uptair to the econd floor, after making everal turn and walking along the hallway, the uer finally topped at the outhwet ide of the building Zhao, 00. In order to judge the ueability of thi data, the dataet wa then evaluated uing a reference trajectory, which wa obtained from backward moothing filter. By comparing it with thi reference trajectory, the navigation olution got an accuracy of 0.5 m mean poition error 3D and 0.7 m maximum poition error 3D. Fig. 4. Field tet trajectory Navigation olution Paageway Navigated Paageway Through mot of the indoor trajectory the GPS wa not available, which lead to the ue of the Wi-Fi data. The receiver had an acce to 50 acce point along the trajectory, however only AP 5 line of ight were ued due to their trong ignal trength. Figure 3 how the navigated trajectory blue, all poible paageway yellow and the proceed navigation olution green. A dicued in the method ection, the proceed navigation olution wa ued epoch by epoch a the input for the propoed map matching algorithm. 45
Mohamed Attia, Adel Moua, ing Zhao, Naer El-Sheimy Figure 5 and 6 how the map matching reult for both algorithm repectively. The figure contain all the poible paageway black, the proceed navigation olution blue and the map matched path red. A hown in the figure 5 and 6, it i clear that the econd algorithm achieved enhanced matching reult on mot of the trajectory, however the econd algorithm didn t get the ame matching reult for the tair area a the firt algorithm did located on the upper right region of the trajectory. A illutrated in table, the firt algorithm achieved a matching percentage of 69.96 % from the navigated paageway, and the econd algorithm achieved a percentage of 7.0 %. Thi percentage meaure the efficiency of the algorithm to detect the correct link and project the olution on thi path. The percentage wa imply calculated by comparing the number of the mied paageway with the total number of the navigated paage way. It can be noticed that although the econd algorithm ha better reult on mot of the trajectory, but ince it didn t achieve the ame reult on the tair region which count a long number of link the overall percentage i not ignificantly higher than the firt algorithm. Thee achieved reult can imply illutrate the ignificance of the propoed algorithm in increaing the reliability of the navigation olution. Thi improvement could be noticed when auming the map itelf a our reference trajectory. The poition error i reduced to a maximum 3D poition error of.3 m and a mean D poition error of 3. m. Navigation olution Map Matched path Building Paageway Fig. 5. Reult for the firt propoed algorithm 46
Aiting peronal poitioning in indoor environment uing map matching Fig. 6. Reult for the econd propoed algorithm Tab.. Matching Reult for both algorithm Navigation olution Map Matched path Building Paageway 5. CONCLUSIONS Correct matched path Firt algorithm 69.96 % Second algorithm 7.0 % In order to enhance the reliability of poitioning olution inide indoor environment, thi paper ue a map matching technique to ait a current pedetrian navigation ytem. The input for the algorithm i the navigation olution obtained from the GPS/INS/Wi-Fi integrated ytem. The algorithm ue the map matching to locate the olution on the mot probable location paageway, increaing the reliability of the overall navigation output. Two matching logic were field teted in the engineering building at Univerity of Calgary, and the matching percentage wa around the 70% for both. The achieved reult of locating the navigated paageway can be very ignificant for many mobile mapping application. Whether to be a tandalone PNS, or built inide a mart cell-phone, thi real-time, low-cot, 47
Mohamed Attia, Adel Moua, ing Zhao, Naer El-Sheimy uable, friendly ytem can be ued in emergencie, law-enforcement, regular indoor navigation, handicapped, blind peron, and many other aiting navigation application. 6. ACKNOWLEDGEMENTS Thi work wa upported in part by reearch fund from TECTERRA Commercialization and Reearch Centre, Canadian Geomatic for Informed Deciion GEOIDE Network Centre of Excellence NCE, and the Natural Science and Engineering Reearch Council of Canada NSERC to Dr. Naer El-Sheimy. Alo, Member of the Mobile Multi-Senor reearch group, the Univerity of Calgary are acknowledged for their effort in field teting. 7. REFERENCES Arto, P., Ari-Heikki, S., Merja, H., Jui, H. and Jonna, H., 009. Toward deigning better map for indoor navigation: experience from a cae tudy. the 8th International Conference on Mobile and Ubiquitou Multimedia, Cambridge, United Kingdom: ACM. Attia, M., Moua, A. and El-Sheimy, N., 00. Bridging Integrated GPS/INS Sytem with Geopatial Model for Car Navigation Application. ION GNSS conference, Portland, USA. Banayake, C., Mezentev, O., Lachapelle, G. and Cannon, M.E., 005. An HSGPS, inertial and map-matching integrated portable vehicular navigation ytem for uninterrupted real-time vehicular navigation. Int. J. Vehicle Information and Communication Sytem, Vol., No. /, pp.3 5. Bullock, J. and Krakiwky, E., 994. Analyi of the ue of digital road map in vehicle navigation. IEEE Sympoium on Poition Location and Navigation, La Vega, NV, pp. 494 50. El-Sheimy, N., 00. Inertial Technique and INS/DGPS Integration. Lecture Note, Geomatic Department, Univerity of Calgary. Gillieron, P. and Merminod, B., 003. Peronal navigation ytem for indoor application. th IAIN World Congre, Berlin, Germany. Glanzer, G., Bernoulli, T., Wieflecker, T. and Walder, U., 009. Semi-autonomou indoor poitioning uing MEMS-baed inertial meaurement unit and building information. 6th Workhop in Poitioning, Navigation and Communication WPNC, pp. 35-39. Greenfeld, J., 00. Matching GPS obervation to location on a digital map. the 8t Annual Meeting of the Tranportation Reearch Board, Wahington D.C.. Khider, M., Kaier, S., Roberton, P. and Angermann, M., 008. The Effect of Map- Enhanced Novel Movement Model on Pedetrian Navigation Performance. th European Navigation Conference ENC GNSS. 48
Aiting peronal poitioning in indoor environment uing map matching Mieenberger, K., Klau, J., Zagler, W., Karhmer, A., Shayeganfar, F., Anjomhoaa, A. and Tjoa, A., 008. A Smart Indoor Navigation Solution Baed on Building Information Model and Google Android. Computer Helping People with Special Need. vol. 505: Springer Berlin / Heidelberg, pp. 050-056. Quddu, M., Ochieng, W. and Noland, R., 007.Current Map Matching Algorithm for Tranport Application: State-of-the art and Future Reearch Direction. Tranportation reearch. Part C, Emerging technologie. Quddu, M., Ochieng, W., Zhao, L. and Noland, R., 003. A general map matching algorithm for tranport telematic application. GPS olution, Volume 7, Number 3, pp. 57-67. Syed, Z., 009. Deign and Implementation Iue of a Portable Navigation Sytem. PhD thei, Department of Geomatic engineering, Univerity of Calgary. White, C., Berntein, D. and Kornhauer, A., 000. Some map matching algorithm for peronal navigation aitant. Tranportation Reearch Part C 8, pp. 9-08. Zhao,., Goodall, C., Syed, Z., Wright, B. and El-Sheimy, N., 00. Wi-Fi Aited Multienor Peronal Navigation Sytem for Indoor Environment. ION ITM conference. USA. 49