857 
PERFORMANCE ASSESSMENT OF A MULTU-SENSOR PERSONAL NAVIGATOR 
SUPPORTED BY AN ADAPTIVE KNOWLEDGE BASED SYSTEM 
D. A. Grejner-Brzezinska a , C. K. Toth b , S. Moafipoor a 
a Satellite Positioning and Inertial Navigation (SPIN) Laboratory, The Ohio State University, USA - (dbrzezinska, 
moafipoor.2)@osu.edu 
b Center for Mapping, The Ohio State University, USA - toth@cfm.ohio-state.edu 
Commission I, ICWG V/I 
KEY WORDS: Navigation, Multisensor Systems, Knowledge Base, Neural Network, Fuzzy Logic. 
ABSTRACT: 
The prototype of a personal navigator to support navigation and tracking of military and rescue ground personnel has been developed 
at The Ohio State University Satellite Positioning and Inertial Navigation (SPIN) Laboratory. This paper provides a review of the 
navigation techniques suitable for personal navigation and follows with design, implementation and performance assessment of the 
system prototype, with a special emphasis on the dead-reckoning (DR) navigation supported by the human locomotion model. An 
adaptive knowledge system (KBS) based on Artificial Neural Networks (ANN) and Fuzzy Logic (FL) has been implemented to 
support this functionality. The KBS is trained a priori using sensory data collected by various operators in various environments 
during the GPS signal reception, and is used to support navigation under GPS-denied conditions. The primary components of the 
human locomotion model are step frequency (SF) and step length (SL). SL is determined by a predictive model derived by the KBS 
during the system’s calibration/training period. SL is correlated with several sensory and environmental data types, such as 
acceleration, acceleration variation, SF, terrain slope, operator’s height, etc. that constitute the input parameters to the KBS system. 
The KBS-predicted SL, together with the heading information provided by the magnetometer and/or gyroscope, supports the DR 
navigation. The current target accuracy of the system is 3-5 m CEP (circular error probable, 50%). A summary of the performance 
analysis in the mixed indoor-outdoor environments, with the special emphasis on the DR performance is provided. 
1. INTRODUCTION AND BACKGROUND 
INFORMATION 
The ability to determine one’s position in absolute or map- 
referenced terms, relative to objects in the environment, and to 
move to a desired destination point is an everyday necessity. 
Recent years brought up an explosion in the development of 
portable devices that support this functionality. A Personal 
Navigation Assistant (PNA) also known as Personal Navigation 
Device (PND) is a portable electronic tool, which combines the 
positioning and navigation capabilities, usually provided by the 
Global Positioning System (GPS), and possibly by other 
navigation sensors. The most commonly used PNAs are the 
hand-held GPS units, which are capable of displaying the user’s 
location on an electronic map backdrop. This generation of 
PNAs (often referred to as first generation PNAs) are primarily 
used in leisure, marine and hiking applications. PNDs first 
entered the market in the early 1980’s, but they were big and 
rather clunky systems that only contained maps of a small area. 
The newest generation of PNDs offers many more features, 
such as real-time traffic information, location of points of 
interest, and utilizes maps of entire continents. They offer 
sophisticated navigation functions and feature a variety of user 
interfaces including maps, tum-by-tum guidance and voice 
instructions that have been developed primarily for car 
navigation. Dead reckoning navigation using data collected by 
sensors attached to the drive train, such as gyroscopes and 
accelerometers, can be used for greater reliability, as GPS signal 
loss and/or multipath can occur due to urban canyons, foliage or 
tunnels. Currently, numerous cellular phone and PDA (Personal 
Digital Assistant) models have GPS-based navigation 
capabilities, aside from their original design as personal 
organizers. 
It should be pointed out here that although the same navigation 
component is used in car and pedestrian navigation, PNDs differ 
from guidance systems for car navigation in many ways 
following from the condition that pedestrians are not tied to a 
road network. Thus, pedestrians are free to use either network 
like systems (walkways or streets) or region-based systems with 
no obvious network structure (parks, train stations, stadiums, 
etc.). Consequently, PNDs providing route commands must go 
beyond network-based navigation- and adapt to the variability of 
the surrounding environments. Hence, the underlying 
framework for the generation of route instructions for pedestrian 
navigation systems is fundamentally different than that of a car 
navigator. An example approach that describes the relation 
between the navigator and path in terms of “topological stages 
of closeness (SOCs), which enable a finer granularity of route 
instructions, and hence, the generation of more accurate route 
instructions” is described in (http://www.i- 
spatialtech.com/whitejpapers/region- 
based_pedestrian_navigation.htm). In general, this task is more 
complicated than its counterpart for the network-dependent 
navigators. However, regardless of network-dependency or 
independency, to guide mobile users along a route, all 
navigators must be able to determine their location in relation to 
the route. Consequently, GPS or any other navigation 
technology must provide a position fix and guiding algorithms, 
needed to determine the current location within the background 
map along the route taken. Any, even the most sophisticated and 
reliable algorithm that matches the position fix with a map will 
not work if there is no position fix. 
1.1 Technologies, Systems and Trends 
In 1999, the Federal Communication Commission (FCC) 
mandated that wireless carriers needed to support delivery of 
location information to 911 operators in the US and that service