Accurate indoor positioning of firefighters using dual foot-mounted inertial sensors and inter-agent ranging

Abstract

A real-time cooperative localization system, utilizing dual foot-mounted low-cost inertial sensors and RF-based inter-agent ranging, has been developed. Scenario-based tests have been performed, using fully-equipped firefighters mimicking a search operation in a partly smoke-filled environment, to evaluate the performance of the TOR (Tactical lOcatoR) system. The performed tests included realistic firefighter movements and inter-agent distances, factors that are crucial in order to provide realistic evaluations of the expected performance in real-world operations. The tests indicate that the TOR system may be able to provide a position accuracy of approximately two to three meters during realistic firefighter operations, with only two smoke diving firefighters and one supervising firefighter within range.

Full text

Accurate Indoor Positioning of Firefighters using
Dual Foot-mounted Inertial Sensors and Inter-agent
Ranging
J.-O. Nilsson, J. Rantakokko, P. Händel, I. Skog,
M. Ohlsson
Dept. of Signal Processing, ACCESS Linnaeus Centre
KTH Royal Institute of Technology
Stockholm, Sweden
K.V.S. Hari
Statistical Signal Processing Lab, Dept. of ECE
Indian Institute of Science (IISc)
Bangalore, India
Abstract — A real-time cooperative localization system,
utilizing dual foot-mounted low-cost inertial sensors and RF-
based inter-agent ranging, has been developed. Scenario-based
tests have been performed, using fully-equipped firefighters
mimicking a search operation in a partly smoke-filled
environment, to evaluate the performance of the TOR (Tactical
lOcatoR) system. The performed tests included realistic
firefighter movements and inter-agent distances, factors that are
crucial in order to provide realistic evaluations of the expected
performance in real-world operations. The tests indicate that the
TOR system may be able to provide a position accuracy of
approximately two to three meters during realistic firefighter
operations, with only two smoke diving firefighters and one
supervising firefighter within range.
Keywords—Firefighter; localization; foot-mounted INS;
ranging; cooperative localization
I. INTRODUCTION
A robust indoor localization system is expected to improve
the safety and efficiency in firefighting operations. The most
important system applications include alarm functionalities,
navigation guidance, and enhanced situation awareness. The
firefighters main concern is their safety, with particular
emphasis on smoke diving (or Breathing Apparatus, BA)
operations. The alarm functionality can reduce the risk for
firefighters being separated, e.g. by implementing an alarm that
activates if the distance between the smoke diving pair increase
above a pre-specified distance. A navigation aid can help
firefighters to quickly withdraw from dangerous areas, or
reduce the time to find and extract firefighters in distress.
Finally, the efficiency in firefighting operations can be
improved with a situation awareness system which includes
information on past positions (trajectories) and rooms/areas
that has previously been searched, thereby avoiding double
searches of these locations.
Surveys of user requirements and localization technologies
are provided in [1-5]. Considering the need for an
infrastructure-free localization system and the stringent size,
weight, power and cost (SWaP-C) requirements, it is believed
that the accuracy and availability requirements can only be
fulfilled by embracing a multisensor fusion approach, utilizing
sensors with complementary error characteristics [1]. Sensors
and localization sub-systems that are being pursued by
different research teams include foot-mounted inertial
navigation systems (INS) [6-7], back-mounted pedestrian dead-
reckoning systems [8-11], magnetometers, barometric sensors
(using a reference sensor at a known height to counter effects
from weather changes), imaging sensors (including visual [12]
and thermal infra-red cameras [13]), Doppler radar [14], radio-
based ranging [15] using synthetic aperture approaches [16],
and cooperative localization approaches [17-18].
A real-time cooperative localization system, utilizing dual
foot-mounted low-cost inertial sensors [19] and inter-agent
ranging, has been presented in [4,20]. Considering typical
deployments and methodology it is conceivable that short-
range ranging transceivers, e.g., based on ultra-wideband
(UWB) transceivers, and cooperative localization techniques
could be applied in firefighter operations. The distances
between the smoke diving pair, and often also to the smoke
diving leader, are short enough to allow accurate inter-agent
ranging. In large incidents the number of firefighters within
range will increase, thereby enabling a higher performance
gain by incorporating cooperative localization.
Scenario-based tests are expected to play an important role
in the performance evaluations of firefighter positioning
systems [21-23], since this ensures that the test subjects use
realistic movements and distances between (and number of)
firefighters. The former affects the performance of the foot-
mounted INS, especially in operations with heavy smoke and
heat that forces the firefighters to move close to the ground.
The latter determines the connectivity between the firefighters
and thus the possibilities of applying cooperative localization
schemes for improved accuracy.
The methodology employed in firefighting operations,
determining the relevant test scenarios, varies between
countries. Using Sweden as an example, a firefighting team
responding to an alarm typically consists of two smoke diving

firefighters, one smoke diving supervisor, one firefighter
responsible for supplying water, and a sector chief. Typically,
two or more of these five-person teams are called in from
different fire stations.
Smoke diving operations are performed with firefighters
working in pairs with short distances between them and whilst
always carrying the water hose. The smoke diving supervisor
will stay in a safe area, yet close enough to the smoke diving
pair so that he can quickly help them if needed. The smoke
diving supervisor should, according to regulations,
continuously keep track of the positions and activities of the
smoke divers; however, in reality it is extremely difficult to
obtain the required information through the radio. The smoke
diving supervisor is equipped with breathing apparatus and
carries his own water hose, enabling him to aid or rescue the
smoke divers if needed. In large incidents, one supervisor can
be responsible for multiple smoke diving teams operating
within the same area, and replacement or emergency rescue
teams can be standing by at the base point prepared for rapid
deployment.
Consequently, in this paper we present initial results from
scenario-based performance evaluations using the system
presented in [4,20]. The evaluations were done with a fully
equipped smoke diving pair and a smoke diving supervisor.
Photographs from the tests are shown in Fig. 1. The results
shows that the positioning system using dual foot-mounted
inertial sensors and inter-agent ranging could potentially
provide the necessary positioning accuracy and availability for
smoke diving operations.
II. DESCRIPTION OF THE TOR SYSTEM
Each firefighter is equipped with a zero-velocity aided INS
[19,24], where tri-axial accelerometer and gyroscope sensors
are integrated into the heel of custom-made shoes. Although
some of the sensor bias errors for foot-mounted INS can be
estimated using zero-velocity updates [25] when the foot is at
stand-still, the position and heading error of a foot-mounted
INS remains unobservable, and grows (slowly) without bound
[4]. Further, the high dynamics of foot-mounted inertial sensors
may cause dynamic dependent errors which further worsen the
situation. Therefore, to reduce the error growth rate of the
individual firefighters (self-contained) positioning modules, the
users are equipped with dual foot-mounted INS (one on each
foot) and the navigation solution of the two systems are
combined. The use of dual foot-mounted INS also increases the
robustness towards crawling and other irregular search
movements.
Since the human body is non-rigid, the relative positions of
the two foot-mounted INS are not fixed, and one cannot
directly relate the navigation solution of one foot-mounted INS
to another. However, there is an upper limit on how far apart
the two foot-mounted INS can be, and the fusion of the two
navigation solutions can be viewed as a filtering problem with
non-linear inequality constraints [26]. Consequently, no
additional measurements are needed to fuse the information
from the inertial sensors mounted on different feet.
A centralized cooperative localization algorithm, utilizing
the information from UWB-based inter-agent ranging devices
combined with the position estimates and uncertainties of each
first responder, has been implemented [20]. However, synthetic
ranging measurements are generated from position
measurements obtained with an Ubisense system (Research &
Development Package), installed in the KTH R1 facility. The
performance of the pre-installed Ubisense system has been
previously evaluated, see [27]. The TOR system is running in
real-time and the firefighter positions are conveyed to all units
every second. The fusion algorithm runs in real-time on an
Android-based processing platform (Samsung Galaxy S3). The
processed data from each firefighter, consisting of position and
heading information, is transmitted to the command and
control system via IEEE 802.11 WLAN radio links.
Note that even though the cooperation is centralized, the
dead reckoning (inertial navigation) is decentralized and hence
the system can operate for significant time periods without
ranging measurements and contact with the central node.
Fig. 1. Typical behavior, movements and firefighter distances during the
scenario-based tests that were performed in the KTH R1 facility.

The architecture of the TOR system is illustrated in Fig. 2
and the hardware components are shown in Fig. 3. The foot-
mounted INS, constrained sensor fusion and cooperative
localization algorithms are described thoroughly in [20].
III. FIREFIGHTER TESTS
Six runs of a simple smoke diving search scenario were
performed in the underground KTH R1 facility, using fully-
equipped firefighters carrying water hose and nosepiece.
Firefighters from two different fire and rescue services were
used in the tests; the firefighters from the different stations
performed the search using somewhat different movements.
The KTH R1 experimental performance space and presence
laboratory [28] resembles an old industrial factory, with a large
open area (over 300 m2), a corridor with rooms, and four
stories with several smaller office rooms on each floor on one
side of the old nuclear reactor hall. A schematic floor-plan of
KTH R1 is shown in Fig. 4, where the approximate positions of
the built-up walls used in the tests are marked with red lines.
In the first part of the scenario the pair of smoke-divers
searched the upper parts colored in light grey in Fig. 4 whilst
walking. They also moved into the corridor to the right and
searched the first left room in the corridor, before they returned
to the starting position. The second part of the scenario
involved searching a heavily smoke-filled area marked with
dark grey in Fig. 4. The latter part of the scenario involved
crawling motions and ascending and descending a spiral
staircase to the second floor. The first and second parts of the
scenario were performed as a consecutive run with the
positioning system. The smoke diving supervisor stayed close
to the starting point during the tests. The duration of each test
was approximately 8-10 minutes.
The position and heading of the firefighters were initialized
towards each other and the map through an initialization
procedure where the firefighters walked a few times along a
line. The cooperative localization algorithm automatically
initializes the relative positions and headings of the firefighters
as described in [29]; however, the map was manually adjusted
to fit the orientation of the firefighters prior to each test.
IV. RESULTS
Fig. 5-7 shows a representative selection of positioning
result from the field tests. Note that all these results are the
results from the real-time measurements, without any post-
processing. The magenta and the cyan trajectories and symbols
indicate the smoke divers and supervisor, respectively. Fig. 5
shows the result from a run when the system worked without
any significant issues. Fig 6 shows a run with the same
scenario but performed by firefighters from a different rescue
service. The search patterns are different but the performance
remains roughly the same. The trajectories follow the true
paths well. A video of the results can be viewed at [30].
In the first part of the test the position accuracy is around
one meter, as exemplified in the corridor where the door
openings were missed by less than a meter. In the second part
of the test, where the firefighters mostly crawled while
searching the area, the maximum position error is estimated to
be between two and three meters. This can be verified for
instance by comparing the horizontal positions when the
firefighters moved up the stairs with the position of the
staircase on the map. Also, the position errors during the
second part of the scenario is partially affected by the errors of
the first part and small position and orientation errors are
expected to remain from the initialization procedure, affecting
the performance throughout the whole runs.
Fig. 2. Illustration of the system architecture of the TOR system, where
each agent is equipped with dual shoe-integrated inertial sensor, a
processing and communications device, and ranging devices.
Fig. 3. TOR hardware. Each agent is equipped with two OpenShoe units
which are connected to a Samsung Galaxy S3 through cables, and
Ubisense radio tags (provides synthetic ranging measurements).
Fig. 4. Schematic of the layout of the KTH R1 facility and the indoor
scenario used during the tests. Red lines indicate synthetic walls that were
built-up prior to the test. The checker-board area indicates the
approximnate start and stop position for the firefighters.

Fig. 5. Test 1. The smoke divers search the built-up rooms by
walking/crawling around the walls of each room. The smoke diving
supervisor is passive and waits at the starting position. The positioning
system provides a position estimate with a few meters of error throughout
the search.
Fig. 6. Test 2. The smoke divers make a quick search of the left and upper
side of the site by only entering each room. Thereafter they perform a more
thoroughly crawling search of the rooms on the first and second floor of the
right side of the site. The smoke diving supervisor follows the smoke divers to
help with pulling the hose.
Fig. 7. Test 3 (right). One of the smoke divers experience issues with his
system and his heading is thrown off resulting in a large error towards the
end. Due to the inconsistency in actual and estimated errors, the system
will only slowly recover.
The TOR system is currently a research prototype which
has not been hardened to withstand firefighter operations. In
several of the tests, one of the firefighters lost connection with
one of his foot-mounted units. This resulted in a somewhat
worse localization performance but the fact that the system still
provides position estimates also highlights the component
redundancy provided by the use of dual foot-mounted units.
Further, in some of the runs, the crawling motion resulted in
lager jumps in the heading. This is likely due to issues with the
dual foot-mounted INS integration. However, the exact cause
of this is not known. The worst example of this is seen in Fig. 7
where the heading of one smoke diver is thrown off towards
the end of the crawling part of the search, giving a large error
following the subsequent movement back to the starting
position. This is also seen to a lesser extent in Fig 6. The
cooperative localization algorithms do not manage to recover
the position error during the end of the tests where range
measurements were available again (after descending the
stairs). This is partially due to the systematic errors not being
captured in the uncertainty (error covariance) estimates of the
individual firefighter sub-systems, which in combination with
robust estimation techniques make the system reject the
position of the specific smoke diver. If accurate range
measurements were available during a longer period, then the
errors would eventually correct themselves slowly.
Two details about the results and setup are worth further
mentioning: the synthetic ranging and the tuning of the sensor
fusion. The synthetic ranging could lead to overly optimistic
results due to avoidance of environmental dependent
systematic errors which would arise from ranging being done

through walls. On the other hand, significant systematic errors
are still present and the ranging shows a worse performance
compared to commercially available ranging solutions [31-33].
Note that the position errors shown in [27] are from ideal near
field conditions (using a horizontal transmitter on a pole) and
without differentiation to get range. Consequently, we believe
the position results obtained by the TOR system in the
evaluations provided herein are not overly optimistic. About
the tuning of the system, since these are the first field tests,
only rudimentary tuning of the sensor fusion has been done.
Especially, no tuning has been done of the foot-mounted
inertial navigation systems to accommodate movements other
than walking. Also, the embedded inertial sensors
(ADIS16367) are of an older generation and better inertial
sensors are readily available. This would suggest that the
performance results are likely on the pessimistic side.
V. CONCLUSIONS
A summary of the position results obtained during
scenario-based tests performed with fully-equipped firefighters
using the TOR system has been presented. The evaluations
indicate that the TOR system could provide sufficiently
accurate position estimates in a future firefighter localization
system. However, further improvements of the mechanical
robustness as well as tuning and adjustment of the sensor
fusion algorithms are needed.
The performance of first responder localization
technologies is strongly affected by scenario-dependent
parameters; hence, scenario-based tests are an important
complement in the performance evaluations of these systems.
VI. FUTURE WORK
First of all, the pre-installed Ubisense system which is
currently used to obtain synthetic range measurements will be
replaced with time-of-flight measuring UWB transceivers.
Both in-house developed UWB transceivers, similar to the
systems described in [34], and the Time Domain P410 are
being integrated into the system. Secondly, the old inertial
sensors are being replaced with wireless and smaller inertial
units. In a second iteration these inertial sensors will be
replaced with sensors integrated into in-soles that can be easily
exchanged when needed [35], as illustrated in Fig. 8. For
information on ongoing development see [36].
These upgrades of the TOR system are expected to improve
the usability and position accuracy considerably, and continued
scenario-based testing, e.g., conducted during regular training
exercises, considering various first responder applications will
be performed to assess the final performance of the upgraded
system.
Finally, exteroceptic sensors, e.g., GNSS-receivers, and the
possibility to use (a few) anchor nodes, will be integrated into
the system and different initialization schemes suitable for first
responder applications will be developed, which perform well
both during favorable GNSS conditions and in GNSS-
challenged environments.
Fig. 8. Illustration of the concept of using removable in-soles equipped with
multiple inertial sensors, a processing capability and a wireless body area
network communication link to the sensor fusion node on the agent.
ACKNOWLEDGMENT
We would like to express our gratitude to the firefighters
from Uppsala Brandförsvar and Brandkåren Attunda who
volunteered to participate in the field trials.
REFERENCES
[1] J. Rantakokko et al, “Accurate and reliable soldier and first responder
indoor positioning: Multisensor syst ems and cooperative localization,”
Wireless Communications Magazine, Vol. 18, pp. 10-18, April 2011.
[2] C. Fuchs et al, “Indoor tracking for mission critical scenarios: a survey,”
Pervasive and Mobile Computing, Vol. 7, pp. 1-15, 2011.
[3] C. Fischer and H. Gellersen, “Location and navigation support for
emergency responders: a survey,” IEEE Pervasive Computing, Vol. 9,
pp. 38-47, 2010.
[4] J.-O. Nilsson, Infrastructure-free Pedestrian localization, PhD
Dissertation, KTH Royal Institute of Technology, Stockholm, Sweden,
2013.
[5] S. Beuregard, Infrastructureless pedestrian positioning, PhD
dissertation, University of Bremen, Germany, 2009.
[6] E. Foxlin, “Pedestrian tracking with shoe-mounted inertial sensors.”
IEEE Computer Graphics and Applications, pp. 38-46,
November/December 2005.
[7] L. Ojeda and J. Borenstein, “Non-GPS navigation for security personnel
and first responders,” Journal Navigation, Vol. 60, pp. 391-407, 2007.
[8] R. Harle, “A survey of indoor inertial positioning systems for
pedestrians,” IEEE Communications Survey Tutorials, pp. 1-13, 2013.
[9] www.trxsystems.com
[10] www.seertechnology.com/naviseer
[11] Y. Ma, W. Soehren, W. Hawkinson and J. Syrstad, “Following the team
into danger,” GPSWorld, June 2013.
[12] M. Veth, Fusion of imaging and inertial sensors for navigation, PhD
dissertation, Air Force Institute of Technology, 2007.
[13] E. Emilsson and J. Rydell, “CHAMELEON on fire – thermal infrared
indoor positioning,” Proc. of Position Location and Navigation
Symposium (PLANS), Monterey, CA, May 2014.
[14] W. Hawkinson, P. Samanant, R. McCroskey and R. Ingvalson,
“GLANSER: Geospatial location, accountability, and navigation system
for emergency responders – system concept and performance
assessment,” Proc. of Position Location and Navigation Symposium
(PLANS), Myrtle Beach, SC, April 2012.
[15] K. Pahlavan, X. Li, and J.-P. Mäkelä, “Indoor geolocation science and
technology,” IEEE Communications Magazine, Vol. 40, 2002.
[16] A. Cavanaugh, M. Lowe, D. Cyganski and R.J. Duckworth, “WPI
precision personnel locator: Inverse synthetic aarray reconciliation
tomography performance,” Proc. of ION International Technical
Meeting (ITM), January 2013.

[17] A. Kealy et al, “Collaborative navigation with ground vehicles and
personal navigators,” Proc. of Indoor Positionind and Indoor Navigation
(IPIN), Sydney, Australia, November 2012.
[18] H. Wymeersch, J. Lien, and M. Win, “Cooperative localization in
wireless networks,” Proceedings of IEEE, 2009.
[19] J.-O. Nilsson, I. Skog, P. Händel, and K.V.S. Hari, “Foot-mounted
inertial navigation for everybody – an open-source embedded
implementation,” Proc. of Position Location and Navigation Symposium
(PLANS), Myrtle Beach, SC, April 2012.
[20] J.-O. Nilsson, D. Zachariah, I. Skog and P. Händel, “Cooperative
localization by dual foot-mounted inertial sensors and inter-agent
ranging,” EURASIP Journal on Advances in Signal Processing, 2013.
[21] J. Rantakokko, E. Emilsson, P. Strömbäck and J. Rydell, “Scenario-
based evaluations of high-accuracy personal positioning systems,” Proc.
of Position Location and Navigation Symposium (PLANS), Myrtle
Beach, SC, April 2012.
[22] J. Rantakokko, Test Methodology for Evaluations of Soldier Positioning
Systems, FOI Report, FOI-R--3819--SE, December 2013.
[23] M. Harris, “The Way Through the Flames,” IEEE Spectrum, vol. 9, pp.
26-31, 2013.
[24] I. Skog, P. Händel, J.-O. Nilsson and J. Rantakokko, ”Zero-velocity
detection – an algorithm evaluation,” IEEE Trans. on Biomedical
Engineering, Vol. 57, pp. 2657-2666, 2010.
[25] C. Jekeli, Inertial Navigation Systems with Geodetic Applications, de
Gruyter, Berlin, 2001.
[26] I. Skog, J.-O. Nilsson, D. Zachariah and P. Händel, “Fusing information
from two navigation systems using an upper bound on their maximum
spatial separation,” Proc. of International Conference on Indoor
Positioning Indoor Navigation (IPIN), Sydney, Australia, November
2012.
[27] A. De Angelis, P. Händel, and J. Rantakokko, “Measurement report:
Laser total station campaign in KTH R1 for Ubisense system accuracy
evaluation,” Tech. Report, KTH Royal Institute of Technology, 2012.
QC20120618.
[28] KTH R1 Experimental Performance Space and Presence Laboratory
www.r1.kth.se
[29] J.-O. Nilsson and P. Händel, “Recursive Bayesian initialization of
localization based on ranging and dead reckoning,” Intelligent Robots
and Systems (IROS), 2013 IEEE/RSJ International Conference on,
Tokyo, Japan, Nov 2013.
[30] http://youtu.be/Qtyq86tJrZ8
[31] www.timedomain.com
[32] D. Dardari, A. Conti, J. Lien and M. Win, “The Effect of Cooperation on
UWB-Based Positioning Systems Using Experimental Data,” EURASIP
Journal on Advances in Signal Processing 2008, 2008:513873
[33] S. Huseth, B. Dewberry, R. McCrosky, “Pulsed-RF Ultrawideband
Ranging for the GLANSER GPS-Denied Emergency Responder
Navigation System,” Proc. of the 2011 International Technical Meeting
of The Institute of Navigation, San Diego, CA, January 2011.
[34] A. De Angelis, S. Dwivedi and P. Händel, “Characterization of a
flexible UWB sensor for indoor localization,” IEEE Trans. On
Instrumentation and Measurements, Vol. 62, pp. 905-913, 2013.
[35] K.V.S. Hari, J.-O. Nilsson, I. Skog, P. Händel, J. Rantakokko and G.V.
Prateek, “A prototype of a first responder indoor localization system,”
Journal of the Indian Institute of Science, Vol. 93:3, pp. 511-520, 2013.
[36] http://www.openshoe.org