Fan Bai (mobility modeling
and IMPORTANT) [graduated with
PhD. Now at GM labs]
[NOTE: A newer version of this page is at this link, please
view the updated version instead ... THANKS!]
This webpage aims to establish a community-wide library of mobile wireless
networks traces and measurements.
The goal is to have the traces, simulation code,
test-suites and models (of mobility, traffic and user behavior)
established by the experts in the field, widely available
for everyone to use and compare against.
Among the major universities that agreed to provide traces
are USC, MIT, UCSD, Dartmouth, UCSB, UIUC, Georgia Tech,
Purdue, UCLA, Rice, Boston U, Columbia,
U Washington, UNC. Links are continuously
If interested and can contribute
Related Work and Publications from the NOMADS group at UFL
* New: Aug., 07*
U. Kumar, N. Yadav, and A. Helmy, "Analyzing Gender-gaps in
Mobile Student Societies," CRAWDAD Workshop poster (colocated
with MOBICOM 2007)
* New: Aug., 07*
J. Kim, Y. Du, M. Chen, and A. Helmy,
"Comparing Mobility and Predictability of VoIP and WLAN Traces,"
CRAWDAD Workshop poster (colocated
with MOBICOM 2007)
* New: July, 07*
U. Kumar, N. Yadav, and A. Helmy, "Gender-based feature
analysis in Campus-wide WLANS," MOBICOM 2007 poster and SRC. [A webpage for the results is available HERE]
* New: July, 07*
W. Hsu, D. Dutta, and A. Helmy, "Profile-cast: Behavior-Aware Mobile Networking," MOBICOM 2007 poster and SRC.
* New: June, 07*
W. Hsu, D. Dutta, and A. Helmy, "Extended abstract: Mining behavioral groups in large wireless LANs," to appear in Proceedings of MOBICOM 2007. [Longer version of technical report available HERE]
In this work we leverage unsupervised learning technique (i.e., clustering) to identify groups of users with distinct behavioral patterns in the WLAN traces. We develop the TRACE framework and use summarized mobility pattern as an example to show the applicability of the framework. We find that university campus is a diverse setting in which hundreds of groups with distinct behavioral modes exist, and the group sizes follow a power-law distribution.
The technique we propose in this paper could be used for better mobility models, behavioral norm establishment and abnormality detection, profile-based services such as advertisement and group-cast, to name a few.
The Time-Variant Community Mobiliy Model is a model we create to capture two important mobility characteristics we observed earlier from WLAN traces. These two
mobility characteristics are skewed location visiting preferences and periodical re-appearance.
While improving the realism of the mobility model, we also keep mathematical tractability as a requirement for the mobility model. We use random-direction mobility
model as the basic building block, modify it to incorporate fore-mentioned mobility chracteristics. We are able to derive two quantities of interest related
to mobility-assisted routing, the hitting time and the meeting time. We intereted in deriving other quantities in the future.
We make the code for our time-variant community model available here. The code has many parameters and provides full flexibility to match with various mobility
scenarios (for full details, refer to the manual). It simulates the hitting time, the meeting time, and prints the movement traces in two option formats:
(1) NS-2 compatible format, or (2) time, location (in x,y coordinates) format.
* New: Oct., 06*
W. Hsu, D. Dutta, and A. Helmy, "Mobicom Poster Abstract: On the Structure of User Association Patterns in Wireless LANs," to appear in Mobile Computing and Communication Review. Earlier version of poster abstract accepted by MOBICOM 2006.
Previous Work and Publications from the NOMADS group at USC
This paper provides the most comprehensive study of WLAN traces to date.
Traces collected from four major universities (~12,000 users) are analyzed
using metrics for individual user and group behaviors. Similarities and
differences across campuses are studied. Conclusions provide great insight
into realistic behavior of wireless users. Most users are 'on' for a small
fraction of the time, number of access points visited (per user) is quite
low, and on-line user mobility is quite low. On average, a user encounters
only 2%-6% of the user population. Encounter-graphs and small worlds are
introduced to model encounter patterns between users. We find that number
of encounters follows a biPareto distribution and the frienship indexes
follow exponential distributions. A paradigm for 'encounter-based
information diffusion' is introduced for efficient data dissemination in
W. Hsu, A. Helmy, "Principal Component Analysis of User Association
Patterns in Wireless LAN Traces", IEEE INFOCOM poster, April
W. Hsu, A. Helmy, "Capturing User Friendship in WLAN Traces", IEEE
INFOCOM poster, April 2006.
Highlights: This study analyzes a class of protocols, MAID, that
utilize mobility for information diffusion.
MAID uses encounter
information to create age gradients towards the target, and can be used
for discovering resources, routing or
locating nodes efficiently in future mobile networks.
Analytical models are developed to evaluate MAID's performance during its
various (transient and steady-state) phases of operation. Extensive
simulations are used to validate these models and to study the
sensitivity of MAID to a rich set of mobility models.
We find that although MAID is sensitive to the mobility pattern, its
steady state performance is, surprisingly, insensitive to velocity.
We identify the properties of the 'age gradient tree' as the key factor to
explain this interplay between mobility and the MAID protocols.
Thanks to the people who have contributed traces and/or encouraged this
Mostafa Ammar, Richard Fujimoto (Georgia Tech),
Kevin Almeroth, Elizabeth Royer (UCSB),
David Kotz, Andrew Campbell (Dartmouth),
Nitin Vaidya, Jennifer Hou (UIUC),
Ness Schroff, Sonia Fahmy (Purdue), Mario Gerla, Medy
Tracy Camp (Colorado School of
Mines), David Wetherall (U. Washington), Victor Bahl (Microsoft
Research), Ed Knightly, David Johnson (Rice),
Rene Cruz (UCSD), Maria Papadopouli, Kevin Jeffay (U North Carolina),
(Columbia), Azer Bestavros, Ibrahim Matta (Boston U), Dina Katabi (MIT),
Stefano Basagni (Northeastern U.), Michele Zorzi (U. Padova/UCSD), Eylem
Ekici (Ohio State U),
Jim Kurose, Brian Levine (U. Mass - Amherst), Srikanth Krishnamurthy,
Michalis Faloutsos (UC Riverside)
This material is based upon work supported in part by the National Science
Foundation under Grant No. 0134650.
Any opinions, findings and conclusions or recomendations expressed in this
material are those of the author(s) and do not necessarily reflect the
views of the National Science Foundation (NSF).
Number of visitors since Jul. 18 2005:
This page was created May '05. Updated continuously...