The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
861 
Figure 1. Personal navigator: sensor configuration. 
At present, the following sensors are used: dual frequency 
Novatel OEM4 GPS receiver with, Honeywell tactical grade 
HG1700 IMU (gyro rate bias ~3-5 °/hr, and accelerometer bias 
of 2.0 mg) impact foot switches used for timing the user’s step 
events, PTB220A barometer (500-1 lOOhPa pressure range, - 
40-140F temperature range, 0.5-10Hz update rate, 0.1-3s 
output averaging time, and 1.5 m height accuracy (1 sigma)) 
and a three-axis Honeywell HMR3000 magnetometer with an 
integrated pitch-roll sensor; up to 20 Hz read-out rate, 1° (level), 
and 2° (tilt) heading accuracy (1 sigma). The GPS carrier phase 
and/or pseudorange measurements in the double difference 
(DD) mode**, undifferenced pseudorange or ionosphere-free 
linear combination of P1P2 pseudoranges, barometric height, 
compass (magnetometer) heading, inclinometer (magnetometer) 
pitch/roll, and the INS-derived position and attitude information 
are integrated together in the tightly coupled Extended Kalman 
Filter with 29 states listed in Tables 2-3. 
Sensor 
Error 
Sources 
Stochastic Error 
Model 
Accelerometer 
Bias 
Random walk 
Scale factor 
Random constant 
Gyroscope 
Bias 
Random walk 
Scale factor 
Random constant 
Barometer 
Bias 
Random walk 
Scale factor 
Random constant 
Magnetometer 
(compass) 
Bias 
Random walk 
Scale factor 
Random constant 
The barometer and compass are introduced to aid height and 
heading estimation, respectively, when GPS signals are blocked. 
These sensors (as well as the IMU and human dynamics model) 
are continuously calibrated when GPS signals are available. 
While forming the theoretical foundations of this multi-sensor 
system and developing the algorithmic concept, an open-ended 
design architecture was considered, which should allow the next 
level of implementation, such as the inclusion of miniaturized 
imaging sensors, e. g., digital and infrared cameras or laser 
range finders. It should also be mentioned that precise timing of 
all sensory data to GPS time is crucial to sensor/data integration. 
Essentially, the GPS time must be externally recovered from 
1PPS (pulse per second) signal, available through a standard 
interface from a GPS receiver. 
2.2 System Design Architecture 
The system’s design architecture is shown in Figure 2, where 
the three primary modes of operation are indicated (1) 
calibration mode, available during the GPS signal reception; it 
represents the initial sensor calibration and KBS 
calibration/training; (2) hybrid navigation mode, when multi 
sensor assembly is used to navigate; since GPS is available, 
continuous sensor and KBS calibration is also performed; and 
(3) DR navigation mode, which kicks in when GPS is blocked. 
A ZUPT static calibration mode is also included that may be 
applied for partial calibration of the IMU sensors if the operator 
may remain stationary for some time period (several seconds to 
a few tens of seconds usually suffice). 
Table 2. Stochastic error models for multi-sensor error sources 
(Grejner-Brzezinska et al., 2007b). 
** This measurement type is of the highest accuracy and 
provides the best calibration results, but requires data 
transfer from a reference base in real time. Pseudorange- 
based stand alone solution is the simplest, but the least 
accurate approach.