2018 International Conference on Indoor Positioning and Indoor Navigation (IPIN), 24-27 September 2018, Nantes, France Corrective Track Form Matching for Real-Time Pedestrian Navigation P. Peltola, F. Seco and A.R. Jiménez LOPSI group/ CAR, CSIC Madrid, Spain {pekka.peltola, fernando.seco, antonio.jimenez}@csic.es T. Moore NGI, The University of Nottingham Nottingham, England terry.moore@nottingham.ac.uk A. Catherall Plextek Ltd. Cambridge, England aled.catherall@plextek.com Abstract—The Global Navigation Satellite Systems (GNSS) receivers often have difficulties within the urban setting. And indoors the ordinary receivers do not function. For a short term gaps in positioning fixing measurements, a foot-mounted pedestrian dead reckoning system can offer a replacement solution. This paper presents a novel corrective solution for the gaps and the low-quality GNSS measurements fusion with the pedestrian dead reckoning (PDR). The corrective track form matcher consists of two parts. A short track form matcher compares the PDR track against the GNSS track and infers anomalies in the GNSS track. The short track form matcher tags the GNSS measurements as reliable or unreliable. The long track form matcher uses the reliable GNSS measurements and transforms the PDR track by weighting these good quality measurements more. The accuracy (error limit of 96%) in the test case improves from 7.56 m to 4.39, this is by 1.72 times. Keywords—track form matching; pedestrian dead reckoning; GNSS; fusion; navigation; track-to-track I. INTRODUCTION Indoor navigation draws an increasing attention. As the technologies develop, the emerging new services for the public will follow. One of the remaining challenges is the development of an accurate navigation system for pedestrians. The current service level for the public can barely offer proximity information in the shopping centers and similar environments. In an urban setting, the GNSS receiver is not capable to follow the user accurately. Especially the multipath and gaps in the positioning measurements are common with the current technologies used in the mobile devices. This paper presents a novel corrective track form matching method to fuse the foot-mounted inertial measurement unit data with the GNSS positioning data. The section II examines the background. The section III introduces the research challenge. The section IV explains the idea of the short track form matching. The section V depicts the idea of the long track form matching. The section VI analyses the test conducted. The section VII concludes the work. II. BACKGROUND The foot-mounted inertial pedestrian dead reckoning (PDR) provides relative position change information. It characteristically has an error of few metres after walking a few tens of meters. This error is dependent on the strapdown quality and the user’s walking style. If the user stops often and shuffles their feet a lot, the harder it is for the dead reckoning system to follow the user. Ju et al. [1] studied the heel-strike and toe-off performance for the foot-mounted pedestrian dead reckoning. The gait phase and the implementation details of the zero velocity updates are important areas that need to be developed to be more accurate. Altun and Barshan [2] further mention that the error produced by the loose mounting of multiple inertial units at chest and the feet should be taken into account. The GNSS receiver functions well in an open area with a good line of sight conditions to the satellites. When surrounded by trees or buildings the multipath and attenuated GNSS signals make the receiver position information unreliable. In addition, the position error information is unreliable. The work of Hsu et al. [3] describes trials made with dead reckoning on mobile phone and the GNSS measurements in the middle of urban canyons. To compensate for these troubles, previous research work has been conducted. The use of heterogeneous, complementary sensors is a necessity for dealing with varying situations arising in urban and indoor scenarios. Pelka et al. [4] aim to create an architecture that is generally applicable for multi-sensor positioning in different situations. The most obvious application of track form matching is within the map matching methods, like Zampella et al. [5] describe. The user trail is matched with the existing map information. Xu et al. [6] used a grid based map to aid the WiFi and magnetic heading sensor navigation system. Aggarwal et al. [7] used a heuristic map matching to compensate for the heading drift. Points on the map are matched with the points from the dead reckoning trajectory. The drift can be eliminated, depending on the map and the road layout. Our application, in this paper is similar strategy of matching, but extended to the matching between the inertial dead reckoning trail and the GNSS trail. In this paper we do not use matching with the map. Copyright Clearance Code