Conference PaperPDF Available

MapUme: Smartphone Localisation as a Service A cloud based architecture for providing indoor localisation services

Authors:

Abstract and Figures

Accurately determining the user's position is con-sidered the key enabling technology for the provision of location based services on smartphone devices. The most promising ap-proaches to date for achieving accurate indoor coverage are based on WiFi fingerprinting in combination with complex estimation and filtering algorithms. In order to meet these computational requirements for a large number of devices a localisation system must be inherently distributed to provide real time responsiveness to all users at all times. Furthermore, to make such a system commercially feasible those computational resources sometimes cannot be maintained by the localisation service provider itself but have to be dynamically allocated in the cloud always based on current and not on peak demand, which depending on the application can vary significantly and thereby render static systems commercially infeasible in these cases. In this paper we will propose such a cloud based architecture enabling the immediate demand driven provision of computational resources to connected devices.
Content may be subject to copyright.
2012 International Conference on Indoor Positioning and Indoor Navigation, 13-15th November 2012
MapUme: Smartphone Localisation as a Service
A cloud based architecture for providing indoor localisation services
Alan McGibney, Christian Beder, and Martin Klepal
Nimbus Centre for Embedded Systems Research
Cork Institute of Technology
Cork, Ireland
Email: {christian.beder,alan.mcgibney,martin.klepal}@cit.ie
Abstract—Accurately determining the user’s position is con-
sidered the key enabling technology for the provision of location
based services on smartphone devices. The most promising ap-
proaches to date for achieving accurate indoor coverage are based
on WiFi fingerprinting in combination with complex estimation
and filtering algorithms. In order to meet these computational
requirements for a large number of devices a localisation system
must be inherently distributed to provide real time responsiveness
to all users at all times. Furthermore, to make such a system
commercially feasible those computational resources sometimes
cannot be maintained by the localisation service provider itself
but have to be dynamically allocated in the cloud always based
on current and not on peak demand, which depending on the
application can vary significantly and thereby render static
systems commercially infeasible in these cases. In this paper
we will propose such a cloud based architecture enabling the
immediate demand driven provision of computational resources
to connected devices.
Index Terms—Smartphone localization, Indoor localization,
Integration platform, Cloud computing
I. INTRODUCTION
Smartphones have long since moved from being a mecha-
nism to access email on the move for business people to an
affordable commodity in the mainstream consumer market.
As a result smartphone devices have woven themselves in
to the fabric of our everyday lives and users expect much
more than simple email access. Driven by user demand many
application developers are creating new and innovative appli-
cations that take advantage of the smartphones capabilities
such as mobile gaming, social networking, news, maps to
name a few. Location-based services plays a key role in
empowering developers and users to take advantage of a
vast array of applications such as geo-social networking,
multimedia guides,targeted advertising and health and safety.
Must of the location based applications are built on the freely
available GPS technology and as a result are restricted to
outdoor scenarios. A significant number of research works
have been published presenting solutions for indoor location
tracking that utilise Wi-Fi fingerprinting in combination with
complex estimation and filtering algorithms [1], [2], [3],
[4], [5], [6], [7]. Although this approach is not without its
challenges [8] it has proven to be the most promising when
providing accurate indoor coverage. However this results in
significant computational requirements to maintain real time
responsiveness to all users at all times for a large number of
devices. Furthermore, depending on the specific application
the demand for location information can vary dramatically
requiring an architecture that is inherently dynamic in nature
when allocating resources. By statically allocating resources
to meet peak demand can render the system commercially
infeasible in some cases.
In this paper we will present MapUme, an opportunis-
tic location system for smart phone devices that enables
localisation services that work seamlessly in heterogeneous
environments including indoors as opposed to GPS based
outdoor-only systems. MapUme is built on a cloud based ar-
chitecture enabling the immediate demand driven provision of
computational resources to connected devices. The presented
indoor localisation system allows the simultaneous accurate
localisation of a large number of users without relying on
heavy client side processing, which would use up the limited
smartphone’s computational and energy resources. We will de-
scribe the proposed architecture and demonstrate the systems
performance under varying peak demand scenarios and show
how commercially feasible smartphone indoor localisation
services can be provided utilising the elastic infrastructure of
the cloud.In particular we will report on our experience in
the trade show application space and the challenges it poses
by requiring the system to provide localisation services to
thousands of exhibitors and visitors over a very short period
of time in a highly dynamic environment.
II. MA PUM E ARCHITECTURE AND COMPONENTS
The advances of Cloud computing brings new opportunities
for location based services. Using a cloud-based architecture,
sensory information is still generated in each client device but
is processed and aggregated within the cloud. This information
is then available for users or services to share among each
other via the cloud or for knowledge mining of these location
datasets across a large numbers of users. The benefits of
cloud computing for location based services are clear, there
is minimum large up-front capital on data centres, eliminates
the need to plan ahead for provisioning and allow companies
to invest in increasing resources as needed [9]. Companies
such as Trimble and Qualcomm offer cloud based solutions
however mostly as a content delivery mechanism rather than
a dedicated positioning service.978-1-4673-1954-6/12/$31.00 c
2012 IEEE
2012 International Conference on Indoor Positioning and Indoor Navigation, 13-15th November 2012
Fig. 1. MapUme General Architecture
MapUme is a computational platform which is built around
the concepts of an event driven architecture combined with
dynamic service composition and invocation principles. The
design goal of the computational platform architecture is to
create a scalable, distributed and extensible platform that
allows for the collecting of sensor data from a large number of
subsystems, to provide computing resources to process those
measurements into meaningful quantities and finally make this
processed information available to business applications on top
of the platform. From the end-users point of view the platform
therefore serves the purposes of providing an abstraction layer
between the sensor/actuators subsystems and the application
that through its scalable and distributed design allows the
leveraging of the processing capabilities of cloud based server
infrastructures. From the developers perspective the platform
offers an architecture that provides interfaces to the devices,
context information and to the business application while at
the same time providing scalability and distributed processing
capabilities.
The MapUme platform comprises of a set of interact-
ing components as depicted in Fig. 1. An important design
paradigm within the platform is inversion of control and
dependency injection. This concept allows the decoupling of
the interface a service provides from its actual implementa-
tion, thereby removing dependencies between the individual
components. Furthermore every service of the platform is
wrapped up automatically within a proxy enabling monitoring
of the service execution as well as a transparent remote
service invocation. The components can be distributed over the
network or into the cloud so that each of those components
runs on an individual machine in order to increase the amount
of processing power necessary to deal with large number
of smartphone devices connected to the system. The plat-
form implements a communications infrastructure provided by
transparent message services. The web-server, message queue
server and database server that are used within the platform
are all off the shelf components and can be interchanged with
other technologies where needed. All these components come
with the ability to run on a cluster, so it can be expected that
they will be able to meet MapUme scalability requirements
without further modifications. The business gateway provides
data and knowledge to end user applications and published via
the web-server. In case the web-server is clustered the business
gateway will run several instances across the cluster, so it
scales with the demand driven from the business perspective.
The platform manager and workers are the key components
of the architecture and will be described below.
A. Platform Manager
The creation of a worker service is usually (but not neces-
sarily) triggered by a device subsystem transmitting its data to
the platform. The platform manager will in this case assign a
service to the device and request it to be created if necessary.
If the appropriate service for the device already exists the data
will be passed on to this service. At the same time broadcast
events in the platform are monitored and relevant messages
are sent back to the device subsystems. The platform manager
is also responsible for monitoring the worker services and
keeping its device to worker service mapping database up to
date. This mechanism allows distributing the worker services
over the network and provides a scalable infrastructure for
meeting extensive processing demand.
B. Workers
The central component of the MapUme platform is the
worker service known as the Localisation System Manager
which is an requestable service within the localisation plat-
form. The task of the Localisation System Manager is to
manage a thread pool of Localisation Engines, one for each
device, and to pass on the sensor data received from the Device
Interface Listener. It receives back the estimated position data
and finally broadcast it through the Position Receiver queue to
all interested services. The Localisation Engine uses the sensor
measurements together with environment data such as floor
plans and fingerprints from an Environment and Configuration
Manager in order to estimate the position of the device. The
MapUme localisation dynamic model is hierarchically struc-
tured and non-deterministic using probabilistic graphical mod-
els (PGMs)[10]. It offers a compact graph-based representation
to model a joint probability distribution of latent and observed
random variables exploiting conditional independencies. These
statistical independence properties are exploited for efficient
inference and learning. Every time the position is updated it is
broadcasted to the Position Receiver queue, which distributes
it to the Logger service that writes everything to the database,
to the Instant Data Generator service, which is responsible for
pushing the updated position back to the devices, as well as to
the Business Gateway. The Business Gateway is then used by
the web server in order to make all the information available
to business applications. The processing intensive parts of
the platform are the localisation engines for each connected
2012 International Conference on Indoor Positioning and Indoor Navigation, 13-15th November 2012
Fig. 2. MapUme Platform for Trade Show Scenario
phone which run within the primary workers. Those workers
are running in the cloud so as to meet this processing de-
mand by allocating additional processing resources if needed.
Certain computational resource intensive tasks could also be
outsourced to the secondary workers by the platform. However
from the scalability point of view it makes no difference if the
primary or the secondary workers are used.
The MapUme computation platform utilises it own cognitive
capabilities to predict and plan the computational requirements
in order to optimally provision computational resources to
the components with the goal of keeping the costs under
control and thereby making the approach feasible in a real-
world commercial application. This context aware processing
approach goes beyond the state of the art of current cloud-
based middleware platforms, which are by design general
purpose processing engines that cannot take any such context
information into account and have to base their load prediction
solely on statistical analysis of past workloads[11], [12].
MapUme’s location services are built on this computation plat-
form for online data collection, aggregation and logging, for
online/offline data processing and analysing and environment
modelling.
III. CAS E STU DY
To analyse the performance of the MapUme technology,
a number of deployments have been carried out in various
sites including an office buildings, college campus, shopping
centres, lab environments and at large trade fair venues. The
most challenging of these environments was the trade fair
scenario, due to the dynamic nature of the environment and
the large number of users that require location and context
information over a short period of time. For large trade show
events visitors wish to utilise their time as efficiently as
possible, exhibitors want to communicate and connect to the
right people, organisers want to provide an experience for all
parties involved that is memorable to ensure repeat business.
The ability to achieve these features is only enhanced by
the availability of real time location and context information
that can be provided by the MapUme platform. Fig. 2 shows
the architecture for the trade fair environment the services
provided include:
Visualisation of your current position in a 3D hall map
and provide navigation to any exhibitor or facility (toilets,
restaurants, etc.)
Locate your colleagues without disturbing them with a
phone call (especially useful for exhibitors)
Exhibitors can add any information they like to their stand
in the map (dynamically and in real time from a web
browser)
Provide targeted information to visitors (suppliers of X
product are highlighted for visitors who have said they
are interested in X product)
The system allows real time communication and updating
for example event/meeting scheduling.
A key requirement from the venue organisers is the ease of
installation and management, the process employed must be
efficient with easy to use tools. The procedure to set up a new
deployment in a trade fair using the MapUme tools takes just
three steps before localisation can take place:
1) Define the map of the environment including the posi-
tion of specific area of interests, i.e. exhibitor booths,
facilities
2) Using a smart phone application calibrate the system by
collecting a fingerprint of the received signal strength in
the area you wish to cover
3) Submit calibration measurement to the server for auto-
mated processing
The remaining steps involve users of the system populating
relevant context information that is communicated to all other
users where appropriate. Fig. 3 shows a screen shot of the
phone application with the users location displayed on the
map of the environment.
Fig. 3. MapUme phone application
The localisation accuracy was measured by collecting an
independent set of calibration data using the fingerprint ap-
plication. This data was then fed into the system using a
device simulator and the deviation between the actual positions
and the positions returned by the system were recorded.
The histogram of these localisation errors in a real trade-fair
environment of approximately 6270m2is shown in Fig. 4.
It can be seen that on average a localisation accuracy of
better than 10m is achievable by the system. The following
sections outline the evaluation of system performance in terms
of system scalability.
To evaluate the performance of the platform in terms of the
number of devices that it can handle workers were deployed on
2012 International Conference on Indoor Positioning and Indoor Navigation, 13-15th November 2012
Fig. 4. Histogram of Localisation Accuracy
Fig. 5. Evaluation of System Scalability
two different machines and the response time was measured,
i.e. the average time between sending an RSSI reading to the
platform and receiving a position estimate back. This was done
for different numbers of devices served simultaneously by the
platform. Fig. 5 plots the response times against the number of
simultaneously served devices. We used an older and slower
machine (.7), which was able to serve up to 10 devices
simultaneously before the response times and their standard
deviation started to deteriorate. On a faster server (.44) the
response times worsened after 30 simultaneous devices were
utilising the platform. However, even with 100 devices the
average response times remain below 10 seconds, although
the increasing standard deviation means that some users have
to wait unacceptably long in this case. To overcome this
limitation one of the major design goals of the platform was
to run distributed and therefore allow for scalability, i.e. allow
serving more devices simultaneously using more servers. It
can be seen in Fig. 5 that if we use both servers together the
number of devices that can be served simultaneously without
any performance degradation increases to 40, which is the sum
of the number of devices served by the individual machines
alone indicating a linear dependence between the number of
devices and the number of servers as desired. Even above that
limit, i.e. when the platform becomes overloaded, the response
times are better if both servers are used together.
IV. CONCLUSION
The future of indoor location based services requires an
architecture that is low cost, dynamic in nature and one
that scales with the varying user demands depending on
the environment utilising these services. This paper presents
such an architecture known as MapUme built on concepts of
distributed and dynamic service composition and invocation.
MapUme was deployed at a large trade fair and was shown
to provide location accuracy within 10m which is more than
sufficient to support the location based services delivered
through its scalable platform for the event. The platform
was designed to meet the challenges posed by the trade fair
environment i.e. the dynamic nature of the environment and
user need for context information.
ACK NOW LE DG ME NT
This work has been supported by Enterprise Ireland though
grant CF/2010/042 and CP/2011/0203.
REFERENCES
[1] F. Seco, A. Jimenez, C. Prieto, J. Roa, and K. Koutsou, “A survey of
mathematical methods for indoor localization,” in IEEE International
Symposium on Intelligent Signal Processing, 2009, pp. 9 –14.
[2] J. Biswas and M. Veloso, “Wifi localization and navigation for au-
tonomous indoor mobile robots,” in IEEE International Conference on
Robotics and Automation (ICRA), 2010, pp. 4379 –4384.
[3] D. Milioris, L. Kriara, A. Papakonstantinou, G. Tzagkarakis,
P. Tsakalides, and M. Papadopouli, “Empirical evaluation of signal-
strength fingerprint positioning in wireless lans,” in Proceedings of
the 13th ACM international conference on Modeling, analysis, and
simulation of wireless and mobile systems, ser. MSWIM ’10, 2010, pp.
5–13.
[4] M. Klepal, M. Weyn, W. Najib, I. Bylemans, S. Wibowo, W. Widyawan,
and B. Hantono, “Ols: opportunistic localization system for smart phones
devices,” in Proceedings of the 1st ACM workshop on Networking,
systems, and applications for mobile handhelds, ser. MobiHeld ’09,
2009, pp. 79–80.
[5] J. Letchner, D. Fox, and A. LaMarca, “Large-scale localization from
wireless signal strength,” in Proceedings of the 20th national conference
on Artificial intelligence, 2005, pp. 15–20.
[6] A. Fink, H. Beikirch, and M. Voss, “Improved indoor localization with
diversity and filtering based on received signal strength measurements,
Computing, vol. 9, pp. 9–15, 2010.
[7] S. Gansemer, U. Gromann, and S. Hakobyan, “Rssi-based euclidean
distance algorithm for indoor positioning adapted for the use in dynam-
ically changing wlan environments and multi-level buildings,” in 2011
International Conference on Indoor Positioning and Indoor Navigation
(IPIN), 2011.
[8] E. Elnahrawy, X. Li, and R. Martin, “The limits of localization using
signal strength: a comparative study,” in First Annual IEEE Communi-
cations Society Conference on Sensor and Ad Hoc Communications and
Networks, 2004, pp. 406 – 414.
[9] Y. Zheng, “Location-based services on the cloud,” 2009, technical
Report, Microsoft Research, MSR-TR-2009-120. [Online]. Available:
http://research.microsoft.com/apps/pubs/?id=102318
[10] D. Koller and N. Friedman, Probabilistic Graphical Models: Principles
and Techniques. MIT Press, 2009.
[11] P. Saripalli, G. V. R. Kiran, R. R. Shankar, H. Narware, and
N. Bindal, “Load prediction and hot spot detection models for
autonomic cloud computing,” in Proceedings of the 2011 Fourth IEEE
International Conference on Utility and Cloud Computing, ser. UCC
’11. Washington, DC, USA: IEEE Computer Society, 2011, pp.
397–402. [Online]. Available: http://dx.doi.org/10.1109/UCC.2011.66
[12] W. Kleiminger, E. Kalyvianaki, and P. Pietzuch, “Balancing load in
stream processing with the cloud,” in Proceedings of the 2011 IEEE 27th
International Conference on Data Engineering Workshops, ser. ICDEW
’11. Washington, DC, USA: IEEE Computer Society, 2011, pp. 16–21.
[Online]. Available: http://dx.doi.org/10.1109/ICDEW.2011.5767653
... The weighted centroid method provided the lowest accuracy; the 5-KNN and the 5-WKNN methods showed an average error rate of 1.5 m; while the Gaussian method outperformed all the others with error rate values under 1 m. The MapUme [11] positioning system is based on a client-server architecture. The client collects the signal strength data and delivers them to the server. ...
Full-text available
Article
Indoor positioning is a recent technology that has gained interest in industry and academia thanks to the promising results of locating objects, people or robots accurately in indoor environments. One of the utilized technologies is based on algorithms that process the Received Signal Strength Indicator (RSSI) in order to infer location information without previous knowledge of the distribution of the Access Points (APs) in the area of interest. This paper presents the design and implementation of an indoor positioning mobile application, which allows users to capture and build their own RSSI maps by off-line training of a set of selected classifiers and using the models generated to obtain the current indoor location of the target device. In an early experimental and design stage, 59 classifiers were evaluated, using data from proposed indoor scenarios. Then, from the tested classifiers in the early stage, only the top-five classifiers were integrated with the proposed mobile indoor positioning, based on the accuracy obtained for the test scenarios. The proposed indoor application achieves high classification rates, above 89%, for at least 10 different locations in indoor environments, where each location has a minimum separation of 0.5 m.
... UbiComp '16, September 12 -16, 2016, HeidelbergSchwaighofer et al. [14] and Ferris et al. [5] leveraged Gaussian process to model RSSI distribution from training data collected beforehand. In 2014, Mcgibney et al. [1] developed a Wi-Fi RSSI and particle-filter-based tracker that achieved a 1.6-m accuracy as the second ranked performance among the state of the art techniques in the Microsoft Indoor Localization Competition 1 . This means that localization systems have already reached sufficient accuracy as a way to recognize human indoor activities. ...
Conference Paper
RSSI-based indoor localization is getting much attention. Thanks to a number of researchers, the localization accuracy has already reached a sufficient level. However, it is still not easy-to-use technology because of its heavy installation cost. When an indoor localization system is installed, it needs to collect RSSI data for training classifiers. Existing techniques need to collect enough data at each location. This is why the installation cost is very heavy. We propose a technique to gather data efficiently by using machine learning techniques. Our proposed algorithm is based on multi-task learning and Bayesian optimization. This algorithm can remove the need to collect data of all location labels and select location labels to acquire new data efficiently. We verify this algorithm by using a Wi-Fi RSSI dataset collected in a building. The empirical results suggest that the algorithm is superior to an existing algorithm applying single-task learning and Active Class Selection.
... In 2014, the Microsoft Indoor Localization Competition 3 was held in conjunction with IPSN2014, where localiza- tion accuracy was used as one of the most important performance criteria. In this competition, a Wi-Fi RSSI and particle filter-based tracker achieved a 1.6-m accu- racy as the second ranked performance among the state of the art techniques, [4] and the results obtained from the submitted methods suggest that accuracy of less than a few meters has not been a challenging issue. The literature in this field has shown that localization performance has been improved continuously during the last decade, thanks to the rapid growth in statistical machine learning research. ...
Article
Indoor localization with wireless communication technologies is gaining popularity among the ubiquitous computing and intelligent robotics communities. Thanks to the rapid growth in statistical machine learning research, indoor localization methods are being actively pursued with accelerated localization accuracy. However, cost in terms of budget constraints with respect to the number of installed sensors should be considered towards practical home sensing, however few studies have addressed this issue. We propose an installation scenario for practical home sensing applications and a flexible procedure for it with optimal sensor selection algorithms. With this scenario, we assume that a large number of ZigBee devices are temporarily deployed to the home in the initial status, then redundant sensors are efficiently taken off via optimal sensor selection. Thanks to a wide variety of options to formulate sensor selection, we describe an empirical evaluation with various formulations. Specifically, we used prototype ZigBee-based wireless sensor systems in an actual residential environment spread over then evaluated the trade-offs between accuracy and the number of sensors. The empirical results suggest that the backward greedy sensor selection algorithm achieves the most stable performance and that a few selected sensors exhibited competitive performance compared to the initial large-number setting.
Full-text available
Technical Report
This deliverable describes work on the creation of occupancy and movement pattern models using mobile phones and other sensory information. It provides a description of the localisation framework implemented for the detection of occupancy in built environments. The occupancy detection is based on an extension and adaptation of the MapUme localisation platform for occupancy detection, developed by CIT and based on sniffing of Wi-Fi signals generated by mobile devices of the occupants and conditioning via wireless sensors for occupancy detection. In specific terms, MapUme consists of the localization system itself as well as a visualization front end for mobile devices. It provides indoor localization based on mobile devices and various other sources of sensory information such as Wi-Fi, Bluetooth and RFID. The MapUme localization technology is being used to contribute localization data to occupancy detection in TRIBUTE. The framework is being extended to enable fusion of context information from various sources to improve the overall accuracy of the occupancy detection. In this document, we present in detail the system that has been deployed for TRIBUTE at the Nimbus Centre (CIT) and is under evaluation for the occupancy level estimation at the room level.
Full-text available
Conference Paper
This document provides a survey of the mathematical methods currently used for position estimation in indoor local positioning systems (LPS), particularly those based on radiofrequency signals. The techniques are grouped into four categories: geometry-based methods, minimization of the cost function, fingerprinting, and Bayesian techniques. Comments on the applicability, requirements, and immunity to nonline-of-sight (NLOS) propagation of the signals of each method are provided.
Full-text available
Conference Paper
This paper presents positioning algorithms for WLAN environments based on RSSI-fingerprinting and Euclidean Distance (EDA). The adapted algorithms can be used in large and dynamically changing environments and multi-level buildings. Evaluation results show a median location estimation error (MLEE) of 2.12m. The discrete third spatial direction (floor level) was correctly estimated in 99% of cases. The calibration effort may be decreased by a factor of 4, if an increase of MLEE error from 2.12m to 2.69m is tolerated.
Full-text available
Conference Paper
Knowledge of the physical locations of mobile devices such as laptops or PDA's is becoming increasingly important with the rise of location-based services such as specialized web search, navigation, and social network applications; further- more, location information is a key foundation for high-level activity inferencing. In this paper we propose a novel tech- nique for accurately estimating the locations of mobile de- vices and their wearers from wireless signal strengths. Our technique estimates time-varying device locations on a spa- tial connectivity graph whose outdoor edges correspond to streets and whose indoor edges represent hallways, staircases, elevators, etc. Use of a hierarchical Bayesian framework for learning a signal strength sensor model allows us not only to achieve higher accuracy than existing approaches, but to over- come many of their limitations. In particular, our technique is able to (1) seamlessly integrate new access points into the model, (2) make use of negative information (not detecting an access point), and (3) bootstrap a sensor model from sparse training data. Experiments demonstrate various properties of our system.
Full-text available
Conference Paper
This paper proposes a novel localization technique based on a multivariate Gaussian modeling of the signal strength measurements collected from several access points (APs) at different locations. It considers a discretized grid-like form of the environment and computes a signature at each cell of the grid. At run time the system compares the signature at the unknown position with the signature of each cell using the Kullback-Leibler Divergence estimation (KLD) between their corresponding probability densities. The paper evaluates the performance of the proposed technique and compares it with other statistical fingerprint-based localization systems. The performance analysis studies were conducted at the premises of a research laboratory and an aquarium under various conditions. Furthermore, the paper evaluates the impact of the number of APs and the size of the measurement datasets.
Article
In the past decades, location-based services have experienced its first generation, in which a user is provided with some specific services in terms of their geo-location. The prosperity of Cloud computing brings new opportunities for LBS and push LBS to a new era. In cloud-based LBS, location information is still generated in each user's mobile client but is aggregated in the cloud. People can share their information among each other via the cloud. Moreover, LBS providers can mine knowledge from these location datasets of large-scale users and better server each user in turn using the knowledge. Overall, the future of LBS is on the cloud.
Conference Paper
Stream processing systems must handle stream data coming from real-time, high-throughput applications, for example in financial trading. Timely processing of streams is important and requires sufficient available resources to achieve high throughput and deliver accurate results. However, static allocation of stream processing resources in terms of machines is inefficient when input streams have significant rate variations-machines remain under-utilised for long periods of average load. We present a combined stream processing system that, as the input stream rate varies, adaptively balances workload between a dedicated local stream processor and a cloud stream processor. This approach only utilises cloud machines when the local stream processor becomes overloaded. We evaluate a prototype system with financial trading data. Our results show that it can adapt effectively to workload variations, while only discarding a small percentage of input data.
Conference Paper
Building upon previous work that demonstrates the effectiveness of WiFi localization information per se, in this paper we contribute a mobile robot that autonomously navigates in indoor environments using WiFi sensory data. We model the world as a WiFi signature map with geometric constraints and introduce a continuous perceptual model of the environment generated from the discrete graph-based WiFi signal strength sampling. We contribute our WiFi localization algorithm which continuously uses the perceptual model to update the robot location in conjunction with its odometry data. We then briefly introduce a navigation approach that robustly uses the WiFi location estimates. We present the results of our exhaustive tests of the WiFi localization independently and in conjunction with the navigation of our custom-built mobile robot in extensive long autonomous runs.
Conference Paper
Cloud computing is leading to transformational changes with stringent requirements on usability, performance and security over very heterogeneous workloads. Their run-time management requires realistic algorithms and techniques for sampling, measurement and characterization for load prediction. Due to the expectation of elasticity, large swings in their demand are common, which cannot be modeled accurately based on raw measures such as the number of session requests, which show very large variability and poor auto-correlation. We demonstrate the use of load prediction algorithms for cloud platforms, using a two-step approach of load trend tracking followed by load prediction, using cubic spline Interpolation, and hotspot detection algorithm for sudden spikes. Such algorithms integrated into the autonomic management framework of a cloud platform can be used to ensure that the SaaS sessions, virtual desktops or VM pools are autonomically provisioned on demand, in an elastic manner. Results indicate that the algorithms are able to match representative SaaS load trends accurately. This approach is suitable to support different load decision systems on cloud platforms with highly variable trends in demand, and is characterized by a moderate computational complexity compatible to run-time decisions.