This document is aim to provide following information regarding the released
dataset:

(1) description of released dataset files structure
(2) introduction about LTE handoff configurations contained in the dataset
(3) method to collect this dataset and reproduce the work on your own

===============================================================================

1. Description of released dataset files structure

The dataset being released is D2 as mentioned in our paper which covers handoff
configurations from 32,000+ cells over 30 carriers in North America, Europe and
Asia.

Both cell level (cell_level_DB_IMC) and sample level (sample_level_DB_IMC)
handoff configurations are released. You can download the tar file directly for
convenience. Sample level dataset contains every instance of handoff
configurations observation during data collection. Cell level dataset only
contains the latest unique handoff configurations.

===============================================================================

2. A brief introduction about LTE handoff configurations contained in the
dataset

In each csv file, the header gives a self-explanatory name. Regarding LTE,
WCDMA and GSM, it is aligned with 3GPP standard. Please refer to 3GPP standards
for detail information.

-------------------------------------------------------------------------------

LTE (Reference: ESTI TS 136.331, 136.304, 136.133)

The first 4 columns in every csv file is MCC, MNC, TAC and Cell ID which can be
used to identify a global unique LTE cell.

allEvents.csv: LTE measurement report trigger events

alloffsetCell.csv: cell specific offset (in LTE measurement object)

alloffsetFreq.csv: frequency specific offset (in LTE measurement object)

allSIB1.csv: System Information Block Type 1, cell-reselection parameters for
evaluating serving cell

allServCell.csv: System Information Block Type 3, cell-reselection parameters
for intra-frequency, inter-frequency and inter-rat.

allSIB4BlackCell.csv: System Information Block Type 4, blacklisted
intra-frequency neighbouring cells

allSIB4NeighCell.csv: System Information Block Type 4, intra-frequency
neighbouring cells with specific cell re-selection parameters

allSIB5.csv: System Information Block Type 5, cell-reselection parameters for
inter-frequency

allSIB5BlackCell.csv: System Information Block Type 5, blacklisted
inter-frequency neighboring cells

allSIB5NeighCell.csv: System Information Block Type 5, inter-frequency
neighbouring cells with specific cell re-selection parameters

allSIB6.csv: System Information Block Type 6, cell-reselection parameters for
inter-rat UTRA

allSIB7.csv: System Information Block Type 7, cell-reselection parameters for
inter-rat GERAN

allSIB8.csv: System Information Block Type 8, cell-reselection parameters for
inter-rat CDMA2000

===============================================================================

3. Method to collect this dataset and reproduce the work on your own

MMLab is a crowdsourcing tool designed to collect handoff configurations at
user device side. It is publicly released as an experiment task of MI-Lab
(http://milab.cs.purdue.edu/) which is an open testbed for in-phone mobile
network experimentation and analytics at scale.

The collected mi2log file contains signaling messages that carry all handoff
configurations in this released dataset.

You can start from the tutorial about how to join MI-Lab and run a specific
task: http://milab.cs.purdue.edu/tutorial_start.

To run the MMLab task, you need to download and install both MI-Lab client app
and MobileInsight app from here: http://milab.cs.purdue.edu/apks/.  In MI-Lab
client app, you should find a task named MMLab and you can use it to conduct
the experiment. Here is some brief explanation about the client app UI:
http://milab.cs.purdue.edu/tutorial_client.