Conference PaperPDF Available

Reliable Geofencing: Assisted Configuration of Proactive Location-based Services

Authors:
  • Technische Universität Berlin, Telekom Innovation Laboratories
Conference Paper

Reliable Geofencing: Assisted Configuration of Proactive Location-based Services

Abstract and Figures

Today, proactive location-based services (LBS) are gaining momentum as mobile devices became able to track the users' position in the background without notable battery drain. They are used in several application areas like location-based marketing or gaming to deliver notifications, e.g. coupons, proactively to the user once a defined geographical area, also known as geofence, is entered. Owning to various reasons like the limited preciseness of positioning techniques for outdoor environments or the strict energy constraints of today's mobile devices, proactive LBS are still facing reliability issues. Either the user is burdened with non-relevant notifications in case the proactive LBS wrongly assumed the user to be located within a geofence, or notifications are not received at all although the user visited the geofence. This paper investigates how the day-by-day configuration of proactive LBS with geofences may influence the reliability of the service as a whole. Based on the findings, it introduces a web service which quantifies how reliable a proactive LBS will most probably be with respect to a given geofence in order to enable even non-experts to properly set up a proactive LBS with appropriate geofences.
Content may be subject to copyright.
Reliable Geofencing: Assisted Configuration of
Proactive Location-based Services
Sandro Rodriguez Garzon and Mustafa Elbehery
Service-centric Networking
Telekom Innovation Laboratories, TU Berlin
Berlin, Germany
sandro.rodriguezgarzon@tu-berlin.de
elbehery@campus.tu-berlin.de
Bersant Deva and Axel Küpper
Service-centric Networking
Telekom Innovation Laboratories, TU Berlin
Berlin, Germany
bersant.deva@tu-berlin.de
axel.kuepper@tu-berlin.de
Abstract—Today, proactive location-based services (LBS) are
gaining momentum as mobile devices became able to track
the users’ position in the background without notable battery
drain. They are used in several application areas like location-
based marketing or gaming to deliver notifications, e.g. coupons,
proactively to the user once a defined geographical area, also
known as geofence, is entered. Owning to various reasons like
the limited preciseness of positioning techniques for outdoor
environments or the strict energy constraints of today’s mobile
devices, proactive LBS are still facing reliability issues. Either
the user is burdened with non-relevant notifications in case the
proactive LBS wrongly assumed the user to be located within
a geofence, or notifications are not received at all although the
user visited the geofence. This paper investigates how the day-by-
day configuration of proactive LBS with geofences may influence
the reliability of the service as a whole. Based on the findings, it
introduces a web service which quantifies how reliable a proactive
LBS will most probably be with respect to a given geofence in
order to enable even non-experts to properly set up a proactive
LBS with appropriate geofences.
I. INTRODUCTION
During the last years, location-based services (LBS) appar-
ently became the most popular type of context-aware service
used on mobile devices. With the introduction of miniaturized
and integrated sensor solutions for continuous and energy-
efficient background tracking on mobile devices, they have
evolved towards a second generation of LBS that are capable
of acting proactively once the user’s location falls into a
predefined region, a mechanism better known as Geofencing
[1]. Nowadays, various domains take advantage of these perva-
sive computing capabilities in order to deliver location-based
coupons to potential customers of brick-and-mortar stores, to
conduct neighborhood-wide polls [2] or to alert citizens in a
smart city if they approach an area of high air pollution [3].
Despite the latest commercial breakthrough of proactive
LBS within different application fields, they are still suffering
from reliability issues [4], [5]. Owing to different technical
limitations, they might miss to detect a user to be within a
dedicated area of interest (geofence) or might burden the user
with non-relevant information once the user already left the
geofence. At first, the accuracy of a position technique comes
into focus [6] since the higher the mean accuracy, the higher
the chance of a proactive LBS to correctly match the user’s
position against a given geofence. Unfortunately, using a high
accuracy positioning technique comes along with an increased
energy consumption [7]. Additionally, in case an architecture
of a proactive LBS comprises mobile devices in combination
with a server infrastructure like in most commercially available
solutions, the communication over the air interface might
become the bottleneck. Moreover, configuring a proactive
LBS with too many, too small, too complex or wrongly-
positioned geofences decreases the reliability of a proactive
LBS significantly [4], [8].
So far, the main focus of research lay onto guaranteeing
a high level of reliability according to a given service level
agreement (SLA) while keeping the energy-consumption and
communication overhead to a minimum. The day-by-day con-
figuration on the other side has received considerably lesser
attention, despite of being vulnerable to human mistakes.
Therefore, the day-by-day configuration with geofences is
conducted at least under the support of an expert in order
to prevent a proactive LBS to be configured in a wrong or
unfavorable way. But to increase the attractiveness of proactive
LBS for non-experts like public authorities [2] or marketing
departments it needs to be ensured that geofences can be
set up easily without an expert supervising it. This becomes
even a strong requirement in case geofences are set up in
a completely automated fashion like in a smart city context
[3]. For that purpose, a web service gets introduced which
determines whether a geofence under investigation fulfills
the reliability-related requirements with respect to a given
SLA. In order to accomplish it, the web service takes the
LBS-specific background tracking properties, the transport
infrastructure and the local characteristics of the environment
into consideration.
II. RE LATE D WORK
In the recent years, all major mobile operating system
providers extended their products with Geofencing features
which can be accessed by mobile application developers via
dedicated API’s. However, the amount of geofences allowed
per application is limited for all these API’s and the geofences
need to be defined in a circular shape. Third-party solutions
like the ESRI Geotrigger Service1as well as academic ap-
proaches [2], [5] try to break these limitations by allowing
to define polygonal shaped as well as unlimited geofences.
The same applies for fully-integrated solutions like offered
by Bitplaces2, Sensewhere3and Plot Projects4, to mention
just a few. While in system-integrated and some third-party
solutions geofences can only be created via API calls, the
fully-integrated solutions usually provide visual editors to
allow non-experts to draw geofences onto a two-dimensional
map like in [2]. Nonetheless, during the use of a visual
editor insights on the type of proactive LBS as well as
the local characteristics of the respective region are highly
recommended for a proper configuration. But so far, none of
the existing commercial nor academic Geofencing solutions
support a geofence designer with a proper estimation on how
a geofence under investigation will most probably affect the
reliability of the proactive LBS, although the configuration has
been proven to influence the reliability [4].
III. CON CE PT
The idea is to support geofence designers, regardless of
being experts or not, during the geofence creation process
with an estimation on how a desired configuration will most
probably affect the overall reliability of a proactive LBS.
For that purpose, a function is realized which validates a
given geofence against various criteria in order to quantify
its influential character onto the perceived reliability. It should
not only support human beings during a manual configuration
process with rough estimations but it should also enable the
introduction of the next generation of LBS that are supposed to
be configured with geofences in a completely automated fash-
ion. These dynamically parametrized LBS will automatically
obtain a huge amount of candidate geofences from various
sources, e.g. environmental sensor readings [3] or driving
schedules [9], and they need to determine for each inferred
geofence whether it negatively affects the reliability of the
proactive LBS, before deploying it.
Generally, a decrease of reliability induced by an adjusted
configuration can have two major reasons: Either the geofences
are formed or located in an inappropriate manner [4] or the
computational costs of matching a position against a huge set
of highly-complex geofences exceeds the given computational
resources. In the latter case, a proactive LBS can limit the
amount of vertices allowed to be defined for a single or all
geometrical geofences, as it is realized within the Geofencing
Extension API5. However, the computational costs will only
be predictable in case the distribution of geofences per region
is known a priori [10]. In the former case, various factors
like the positioning technique in use, the underlying road
network and the local radio signal characteristics need to
be considered within a validation function to make proper
1https://developers.arcgis.com/geotrigger-service/
2https://www.bitplaces.com
3http://sensewhere.com
4http://www.plotprojects.com
5https://developer.here.com
assertions about the geofences’ influence on the reliability of
the proactive LBS. In the following, the most influential factors
are enumerated and discussed.
A. Parameters of proactive LBS
The most influential factor originating from the proactive
LBS itself is called hover time [6]. It is defined as the
maximum time span a mobile device needs to be located
within a geofence, so that the proactive LBS will be able
to detect it. It results directly from the sampling rate of the
positioning technique which is the number of times a position
is measured and calculated within a given time unit. If a
mobile user is moving fast and the geofence is comparable
small it might happen that the user transits the geofence
without the positioning logic taking notice of it. On the other
side, if a geofence is big enough, it takes for a mobile user
at least the hover time to transit the geofence and at least
one position fix will fall into the geofence. In the best case,
the hover time is small enough so that even a very small
geofence of a few meters of diameter will not be overseen by
the positioning logic. Unfortunately, the higher the sampling
rate, the higher is the energy consumption of a positioning
technique. As a consequence, various approaches try to find
the optimal balance between the required energy-consumption
and a minimum geofence diameter for reliable Geofencing.
Each deployment of a proactive LBS comes along with their
own hover times as stated within the accompanying SLA’s.
Formerly, one single and fixed sampling rate was defined,
regardless of the activity-dependent velocity of a user. Since
the emerge of energy-efficient activity recognition algorithms
for mobile devices, the sampling rate gets dynamically ad-
justed to the activity of a user [5], [11]. Hence, rather than
considering a single sampling rate during the validation of a
geofence, the envisioned function needs to deal with activity-
dependent sampling rates. Moreover, waking up a mobile
device from standby or sleep mode at a fixed interval to initiate
the positioning process is draining the battery significantly.
Therefore, mobile operating systems schedule the next wake
up time based on the requirements of various background
services in order to reactivate the processing unit for more than
one background service at a time. To enable this flexibility,
the sampling rate for a background service is defined as a
minimum and a maximum time span in which the next wake
up needs to happen.
B. Local infrastructure
Nonetheless, the aforementioned intrinsic parameters of a
proactive LBS and the form and size of a geofence under
investigation are not sufficient information to make a proper
rating for the geofence. It is so far not clear how a user
most probably transits the geofence. The user might just move
alongside the edge of the geofence entering only for a short
moment or might pass it along the longest possible transit
path. The only way to determine possible transit paths is to
take the location and orientation of the geofence and the road
and path network into consideration. Given that, it allows a
validation function to identify all possible paths through the
geofence for all types of movement activities, e.g. pedestrians
or drivers. Assuming a user to move with a constant and
activity-dependent speed, it can then be calculated for all
activities of how probable it will be to receive a proactive
notification with respect to that geofence. The validation
function would even benefit from additional information about
the speed limits and infrastructural obstacles like traffic lights
and road toll plazas which are supposed to have an influence
on the time a user spends onto a certain path through the
geofence.
With the help of the infrastructural knowledge, a validation
function can make first rough estimations about the expected
reliability for a geofence per activity. But so far, all paths are
treated equally, even though some are used more frequently
than others, e.g. more people are expected to use the main
streets instead of a side road per day. This imbalance can
be considered within the validation function by putting more
weight onto the estimations for the most probable path which
may include a highway for example. However, these weights
are still assumptions which of course need to be backed up by
real-world experiences. Hence, taking the local dynamics of
an environment like the daily citizen flows into consideration
would make the estimation even more precise. Other dynamics
of interest are the regular traffic conditions at certain hot spots
like the daily traffic jams which can be obtained via third-party
services.
C. Local characteristics of radio signals
Although GPS is supposed to be the most accurate out-
door positioning method used today, it still suffers from a
significant decrease of accuracy in certain situations. The time-
dependent reasons can be manifold ranging from a bad satellite
constellation to ionospheric interferences. But at some places
in cities GPS is constantly experienced with a low accuracy
because high-rise buildings are obscuring the line of sight
to the satellites. At these locations, a geofence with too fine
grained structures is of disadvantage since the positioning error
is too high. For example, the user might be wrongly located
inside of a nearby geofence although she/he is just closely
passing it. Given the average accuracy of GPS at a particular
location, it is possible to validate if a fine grained structure
of a geofence is overlapping with an area of low average
GPS accuracy. If it’s the case, the estimated reliability of the
geofence can be reduced by a certain amount which depends
upon the average GPS accuracy at the border region as well
as the dimension of the border region compared to the overall
area covered by the geofence.
Recently, mobile devices started to continuously measure
WiFi and cellular radio signals in order to position itself more
energy-efficiently at the expense of accuracy. The monitored
radio signals will be compared to known location-annotated
radio signal fingerprints, matched against known network iden-
tifiers (cellID) or used to triangulate the position by calculating
the distance to at least three known WiFi access points or cell
towers via time-of-arrival or path loss. Since most geofences
π
ϕ
a) b) c)
π
π1
π2
ϕϕ
< -84 dBm
< -84 dBm
Fig. 1. Area marked with dotted lines in which a proactive LBS detects a
mobile to be within the geofence φwhile using the cellID (a) or radio signal
fingerprinting and triangulation methods (b) with a directed antenna π1(c).
are created as circular- or polygonal-shaped areas, independent
of the positioning technique in use, it needs to be verified
whether the given WiFi and cellular infrastructure permits a
mobile device to properly detect enter and leave events in case
WiFi and cellular positioning is used. Figure 1 a) shows an
exemplary rectangular-shaped geofence φwith an idealized
coverage area of a WiFi network πin its vicinity. Assuming
a mobile device to use the cellID method for positioning, it
can only determine whether it is located inside or outside of
the WiFi coverage area by either sensing the WiFi network or
not. Therefore, a proactive LBS can not distinguish whether
the mobile is located in the subareas φπor π\φ. If the
geofence is associated with the WiFi network, a proactive
LBS will detect an enter event as soon as it senses the WiFi
network. Hence, sometimes an enter event is detected although
the mobile device is not located within the geofence (false
positives in π\φ). In the same case, it will not detect an
enter for the subarea φ\πcause it is not overlapping the
WiFi coverage area (false negative). The ratio between false
positives, false negatives and true positives (detection of enter
event while being located in φπ) can be estimated in advance
by investigating all three subareas separately with respect to
the various factors mentioned before and by putting the results
in relation to each other. The same applies for radio signal
fingerprinting or triangulation methods. However, the numbers
of false positives is slightly reduced in these cases because the
subarea π\φis smaller due to stricter constraints, e.g. limiting
this subarea to the coverage of < -85dBm (see Figure 1 b) and
c)).
IV. IMPLEMENTATION
As a proof of concept, the validation function was imple-
mented as an ordinary web service within the infrastructure
of a telecommunication operator instead of a build-in function
within a mobile operating system because processing a request
can take up a considerable amount of CPU power and is
based on a huge amount of infrastructural data acquired from
different sources. It can be accessed via a RESTful API. For
a proper estimation, the web service needs to be parametrized
with the geofence geometry in WKT format, the intrinsic
parameters of the proactive LBS, the targeted reliability and a
flag defining the envisioned use case. The latter parameter is
used by the validation function to make estimations based on
the application scenario, e.g. proactive LBS for pedestrians.
The infrastructural knowledge comprising the road network
a)
P(Estimation): 0.608-0.956
P(Evaluation): 0.416
P(Estimation): 0.827-1.000
P(Evaluation): 0.732
b)
Fig. 2. Estimated (min, max) and evaluated probabilities for two sample
geofences. The simple squared geofence (a) shows a lower estimated and
evaluated probability than the optimized geofence (b).
is provided by the OpenStreetMap project and stored within
a PostgreSQL database with the PostGIS extension for opti-
mized spatial queries. The average GPS accuracy at a certain
location is acquired via a service provided by the CATLES
project [12] while the WiFi and cell networks are given by
the OpenMobileNetwork6. Currently, the web service does
not take the WiFi and cell infrastructure into consideration.
After processing a request, the web service responses with a
list of estimated probabilities for each activity describing the
likelihood for the proactive LBS to correctly detect a geofence
enter event. Additionally, an overall rating is given based on
the application scenario.
V. EVAL UATION
The validation function was evaluated in order to verify
the applicability of the approach. Therefore, two geofences
were created at a T-shaped neighborhood corner within the
proactive LBS of the FlashPoll ecosystem [2]. As shown
in Figure 2, the two geofences represent a simple squared
geofence and another geofence optimized with the validation
function. The evaluation was executed by using the CATLES
simulator, which was used to emulate the way through as
a pedestrian (at 5km/h) at the neighborhood corner. The
detection of the two geofences has been evaluated with a set
of five mobile devices running Android and sampling rates
between 61 and 104 seconds. For each geofence a total of
60 evaluation way throughs were conducted. The evaluation
shows that higher estimated probability values provided by the
validation function also imply a higher detection rate during
the evaluation with CATLES.
VI. CONCLUSION AND FUTURE WO RK
This work presented a method to assist experts and non-
experts during the configuration of proactive LBS with rough
estimations of how a geofence under investigation will most
probably impact the reliability of a proactive LBS. Several
environmental conditions including the road and path network,
the local characteristics of radio signals as well as the intrinsic
parameters of a proactive LBS itself were identified to have an
influence onto the likelihood of a proactive LBS to correctly
detect a user to be within a given geofence. As a proof of
concept, a prototypical validation function for geofences was
6http://www.openmobilenetwork.org/
implemented as a web service and evaluated with respect to
whether the estimated probabilities are good approximations
of the reliability. The web service is of practical use for
various proactive LBS implementations since the estimations
are calculated independently of the given architecture and
are thus generally applicable. Although the prototypical web
service shows good estimations, there is still space left for
improvements. Other environmental conditions like traffic
lights, recurring traffic jams, the population density or even
the demographics of a district might not only improve the
estimations in the first two cases but would also enable
to calculate a rating based on a more detailed description
of the envisioned application scenario, e.g. targeting young
people. Rather than just estimating the geofences’ impact
on the reliability, a future version of the web service might
also propose modified versions of a geofence’s geometry,
orientation or location for an improved configuration.
REFERENCES
[1] U. Bareth, A. Küpper, and B. Freese, “Geofencing and Background
Tracking - The Next Features in LBS,” in Proc. of the 41th Annual
Conf. of the Gesellschaft für Informatik e.V. (INFORMATIK 2011), vol.
192. Berlin, Germany: Köllen Druck + Verlag GmbH, Oct 2011.
[2] B. Deva, S. Rodriguez Garzon, and A. Küpper, “FlashPoll: A Context-
aware Polling Ecosystem for Mobile Participation,” in Proc. of 19th
Conf. on Innovations in Clouds, Internet and Networks(ICIN 2016), Mar
2016, pp. 169–176.
[3] A. M. Muriach, “INFORMATION PROVISION IMPROVEMENT
WITH A GEOFENCING EVENT-BASED SYSTEM,” Master’s thesis,
Universidade Nova de Lisboa, 2015.
[4] M. Alsaqer, B. Hilton, T. Horan, and O. Aboulola, “Performance Assess-
ment of Geo-triggering in Small Geo-fences: Accuracy, Reliability, and
Battery Drain in Different Tracking Profiles and Trigger Directions,”
Procedia Engineering, vol. 107, pp. 337 – 348, 2015, humanitarian
Technology: Science, Systems and Global Impact 2015, HumTech2015.
[5] S. Rodriguez Garzon, B. Deva, G. Pilz, and S. Medack, “Infrastructure-
assisted Geofencing: Proactive Location-based Services with Thin Mo-
bile Clients and Smart Servers,” in Proc. of the 3rd IEEE Int. Conf. on
Mobile Cloud Computing, Services and Engineering. IEEE Computer
Society, 2015.
[6] A. Greenwald, G. Hampel, C. Phadke, and V. Poosala, “An Econom-
ically Viable Solution to Geofencing for Mass-Market Applications,
Bell Labs Technical Journal, vol. 16, no. 2, pp. 21–38, Sep 2011.
[7] U. Bareth and A. Küpper, “Energy-Efficient Position Tracking in Proac-
tive Location-Based Services for Smartphone Environments,” in Proc. of
the IEEE 35th Annual Computer Software and Applications Conference
(COMPSAC 2011). IEEE, July 2011, pp. 516–521.
[8] S. Rodriguez Garzon and B. Deva, “On the Evaluation of Proactive
Location-Based Services,” in IEEE 39th Annual Computer Software and
Applications Conf. (COMPSAC), vol. 2, 2015, pp. 585–594.
[9] A. Nait-Sidi-Moh, W. Ait-Cheik-Bihi, M. Bakhouya, J. Gaber, and
M. Wack, “On the Use of Location-Based Services and Geofencing
Concepts for Safety and Road Transport Efficiency,” in Proc. of the
2013 Int. Workshop on Mobile Web Information Systems (MobiWIS),
M. Matera and G. Rossi, Eds. Springer Int. Publish., 2013, pp. 135–
144.
[10] F. Cirillo, T. Jacobs, M. Martin, and P. Szczytowski, “Large Scale
Indexing of Geofences,” in 2014 Fifth Int. Conf. on Computing for
Geospatial Research and Application (COM.Geo), Aug 2014, pp. 1–8.
[11] T. Yu-Han Chen, A. Sivaraman, S. Das, L. Ravindranath, and H. Bal-
akrishnan, “Designing a context-sensitive context detection service for
mobile devices,” MIT Computer Science and Artificial Intelligence
Laboratory, Tech. Rep., Sep 2015.
[12] S. Rodriguez Garzon, B. Deva, B. Hanotte, and A. Küpper, “CATLES: A
Crowdsensing-supported Interactive World-scale Environment Simulator
for Context-aware Systems,” in Proc. of the 2006 IEEE/ACM Int. Conf.
on Mobile Software Engineering and Systems (MobileSoft), May 2016,
pp. 77–87.
... The reliability of the old concept of geofencing by a static perimeter is in some cases questionable [21]. In the case of geofencing perimeters, the possibility remains that some other objects of interest are inside the perimeter and therefor create false switch-on events. ...
Conference Paper
Full-text available
With the rise of smart mobile devices, context-aware mobile applications became an integral part of our everydaylife. However, testing, evaluating or demonstrating context-aware mobile applications for outdoor environments remains a resource-intensive and cumbersome task. The investigations need to be conducted in real-world experiments in order to gain proper insights into how well an application will perform in a particular environment with its distinct contextual properties. Although steps were undertaken to simulate outdoor environments within the lab, these approaches are either based on radio propagation models for the simulation of wireless networks or are limited to a particular virtual space. To overcome these shortcomings, the outdoor environment simulator CATLES is introduced. It allows to transparently simulate the position of a mobile device by interactively controlling an avatar within a virtual 3D representation of the world. At the same time, it makes use of crowdsensed publicly available WiFi and cell measurements to properly simulate the environment the context-aware mobile application residing on the respective mobile device is supposed to sense at the simulated position.
Conference Paper
Full-text available
The rise of smart mobile devices created new opportunities for electronic and mobile participation. These devices provide means to sense the current environment citizens are situated in, which can be then used to carry out targeted opinion polls. The context-aware polling system FlashPoll takes advantage of this notion and enables the conduct of short and fast context-specific polls which are matched against the citizens’ individual situation. Thus, the target groups are more focused, which increases the relevance of polls and results to the citizens and the executing authorities. This work presents a complete overview of the FlashPoll ecosystem, which comprises poll management tools for public authorities and mobile applications for participants. It has been tested in several field trials to prove the applicability of the approach.
Conference Paper
Full-text available
Advanced systems to assist drivers and increase road safety and efficiency are developed in recent years using new technologies, such as navigation systems, embedded systems, communication technologies, and web-based applications. However, the implementation of new solutions using these technologies requires concepts based on information and resources sharing and inter-services communication. Sharing information between services allows acquiring more knowledge about the current events on the road that helps participants making appropriate decisions. The paper introduces a service-oriented platform integrating Geofencing techniques for real-time tracking of mobile devises. A prototype was developed and deployed within the EU project ASSET-Road1, and preliminary experiment results are reported to show the feasibility and the impact of this solution on safety and road transport efficiency.
Article
Full-text available
Geo-fencing has been predicted to be a multi-billion dollar market inareas such as retail, ambient intelligence, entertainment, healthcare, etc. Businesses have been adopting geo-fencing technology, and now there are several platform providers such as Google, Qualcomm, Esri, Urban Airship, and others. These tools are continuing to attract application developers; however, best practices for choosing the specific performance options within this technology is still ambiguous. For example, Esri provides a geo-trigger service that allows developers to send targeted messages to users when they enter, exit, or dwell in a geo-fenced area. This service also provides the ability to choose higher levels of accuracy or battery saving by offering different location tracking profiles. This paper investigated two geo-trigger tracking profiles (Fine and Adaptive) to assess their performance in small, outdoor, geo-fenced areas; these two profiles are the most accurate but vary in their battery-use. The results show the Adaptive tracking profile to provide 100% reliability and average accuracy of 68.53 meters in geo-fences between 20-70 meter radii. In addition, the Adaptive tracking profile saved 15.20% battery-life while the user is stationery and 9.23% while the user is moving.
Article
Full-text available
This paper describes the design, implementation, and evaluation of Amoeba, a context-sensitive context detection service for mobile devices. Amoeba exports an API that allows a client to express interest in one or more context types (activity, indoor/outdoor, and entry/exit to/from named regions), subscribe to specific modes within each context (e.g., "walking" or "running", but no other activity), and specify a response latency (i.e., how often the client is notified). Each context has a detector that returns its estimate of the mode. The detectors take both the desired subscriptions and the current context detection into account, adjusting both the types of sensors and the sampling rates to achieve high accuracy and low energy consumption. We have implemented Amoeba on Android. Experiments with Amoeba on 45+ hours of data show that our activity detector achieves an accuracy between 92% and 99%, outperforming previous proposals like UCLA* (59%), EEMSS (82%) and SociableSense (72%), while consuming 4 to 6× less energy.
Conference Paper
Full-text available
During the last years, location-based services (LBS) became one of the most popular features of mobile devices like smartphones or tablets. Basically, reactive LBS provide means to access location-specific content upon request by the user, whereas their proactive counterparts are used to push location-specific content to the user as soon as a dedicated zone-called geofence-is entered or left. The fields of application for proactive LBS are manifold, ranging from location-based reminding over location-based citizen participation to location-based marketing. However, evaluating and comparing proactive LBS regarding their performance and scalability are still cumbersome tasks, involving resource-intensive field trials and in-depth investigations of their underlying architectures. This paper identifies domain-and architecture-independent attributes of proactive LBS that are well-suited to characterize a proactive LBS regarding these non-functional properties. It also takes a deeper look into the key parameters which are mainly affecting the performance and scalability in the real world. While putting the focus on these key parameters and attributes, it is proposed to simulate a proactive LBS within an in-lab environment prior to its evaluation in the wild. For this purpose, a prototypical in-lab evaluation framework for proactive LBS is introduced which generates simulation-based reports regarding the functionality, performance and scalability of a proactive LBS in general.
Conference Paper
Full-text available
With the increasing popularity of smartphones, location-based services became a hot topic and a great number of solutions were presented over the last years. The majority of applications are based on the idea to present location-specific information in case the smartphone user asks for it. A relatively small amount of applications are dealing with geo-notifications that are intended to inform the smartphone user proactively about location-specific information in case a dedicated zone is entered or left. The technology behind proactive location-based services is called Geofencing and it is mainly implemented and executed at the mobile device. This paper presents a new approach to offload this resource intensive process of monitoring the user's location into the infrastructure. The mobile device is thereby considered to be a thin client that is mainly responsible to locate itself whereas the continuous comparison of the mobile's position with a large set of dedicated zones, called geofences, is executed within an environment with lower resource constraints. A prototypical implementation of a thin client as well as the corresponding location processing unit within the infrastructure gets introduced, discussed and evaluated under different environmental conditions.
Conference Paper
The modern smart phone and car concepts provide a fertile ground for new location-aware applications, ranging from traffic management to social services. While the functionality is partly implemented at the mobile terminal, there is a rising need for efficient backend processing of high-volume, high update rate location streams. It is in this environment that geofencing, the detection of objects traversing virtual fences, is becoming a universal primitive required by an ever-growing number of applications. To satisfy the functionality and performance requirements of large-scale geofencing applications, we present in this work a backend system for indexing massive quantities of mobile objects and geofences. Our system runs on a cluster of servers, achieving a throughput of location updates that scales linearly with number of machines. The key ingredients to achieve a high performance are a specialized spatial index, a dynamic caching mechanism, and a load-sharing principle that reduces communication overhead to a minimum and enables a shared-nothing architecture. The throughput of the spatial index as well as the performance of the overall system are demonstrated by experiments using simulations of large-scale geofencing applications.
Conference Paper
Location-based services (LBSs) gain a massive shift in popularity these days and are about to take the next step towards proactive LBSs. In comparison to conventional LBSs, their proactive variant needs continuous position tracking to monitor spatial objects to detect the relationships between a user and surrounding objects and proactively perform actions. But tracking so far also results in severe battery drain in mobile devices due to deficient positioning APIs and the absence of energy-efficient positioning methods with adequate accuracy. By exploiting the combined information from several positioning technologies with different characteristics in terms of energy consumption, accuracy, precision and availability, this paper proposes a hierarchical positioning algorithm, which provides a general algorithmic optimization in order to extend existing positioning APIs to energy-efficiently track a user's position without diminishing accuracy. The algorithm dynamically deactivates different positioning technologies and only activates the positioning method with the least energy consumption that at the same time provides sufficient accuracy to correctly determine topological relationships. In that way, the algorithm can reliably and accurately determine, if the user leaves or enters predefined geographic areas to trigger events for proactive LBSs while preserving valuable energy resources. First results on Android handsets show a reduction in energy consumption of up to 90 percent in comparison to conventional GPS tracking.
Article
Geofencing services deliver location-relevant information to mobile subscribers, who enter a geographic “fence” or boundary around the information's demarcation area. With Geographic Positioning System (GPS) capabilities on many phones, such services facilitate a variety of mass-market applications ranging from mobile proximity marketing to proximity-based dating. Applications delivering such services operate on top of a geofencing engine that manages location tracking of subscribers and returns triggers when fence-crossing events occur. We discuss the critical aspects of geofencing and the challenges faced when deploying such services to the mass market. We present an economically viable geofencing solution that scales to large populations and supports a high number of fences per subscriber. Through distributed processing supported by an appropriate client-server protocol, this solution optimizes the air interface usage and mobile battery power. The technical viability of our solution is supported by actual traffic data from mobile proximity marketing and a social networking service. © 2011 Alcatel-Lucent. © 2011 Wiley Periodicals, Inc.