[Dataset] sunysb/mobisteer (v. 2007-06-30)

top

the initial version

@MISC{sunysb-mobisteer-2007-06-30,
  author = {Vishnu Navda and Anand Prabhu Subramanian and Kannan Dhanasekaran and Andreas Timm-Giel and Samir R. Das},
  title = {{CRAWDAD} data set sunysb/mobisteer (v. 2007-06-30)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/sunysb/mobisteer},
  month = jun,  
  year = 2007
}
					

This data set includes data traces that were collected from a moving car 
equipped with an electronically steerable directinal antenna. We drove 
the car in two different environments in Stony Brook University campus 
- Apartment Complex and Parking lot.

We investigate the use of directional antennas and beam steering techniques 
to improve performance of 802.11 links in the context of communication between 
a moving vehicle and roadside APs. To this end, we develop a framework called 
MobiSteer that provides practical approaches to perform beam steering. 
MobiSteer can operate in two modes - cached mode - where it uses prior radio
survey data collected during "idle" drives, and online mode, where it uses probing. 
The goal is to select the best AP and beam combination at each point along the 
drive given the available information, so that the throughput can be maximized. 
We conducted extensive experiments and collected data traces using a commercially 
available eight element phased-array antenna.

Our directional antenna set up uses electronically steerable Phocus Array 
antennas from Fidelity Comtech for the 2.4 GHz band used in IEEE 802.11b/g. 
The Phocus Array antenna system consists of eight element phased arrays
driven by eight individual T/R (transmit-receive) boards that receive radio 
signals from the wireless card via an eight way RF splitter. The phased arrays 
combine radio waves by introducing different phase differences and gains 
in the eight arrays. A T/R board is essentially a vector modulator with 
bi-directional amplifier controlled by software. Various beam patterns are 
possible by setting the phases and gains in different boards. The antenna is
set to behave identically for transmit and receive, i.e., the antenna gains 
for transmit and receive are the same. 

A software program running on an embedded computer (a Soekris net4511 
board) controls the antenna over a serial-line interface to produce 
different beam patterns. The embedded computer is equipped with a 
802.11 b/g miniPCI card based on Atheros chipset with the external 
antenna interface. The embedded computer runs pebble Linux with the Linux 
2.4.26 kernel and the widely used madwifi device driver for the 802.11 interface. 

For convenience, we will refer to the entire vehicular setup, including the 
embedded computer with 802.11 and GPS interfaces and the directional antenna 
as the MobiSteer node.

The APs are Soekris net4826 router boards with similar Atheros based 802.11 a/b/g 
miniPCI cards connected to regular rubber duck omnidirectional antennas. The APs 
also run the same base software (pebble Linux and madwifi driver) as the MobiSteer node.

The APs operate in pseudo-ad hoc mode and continuously unicast 1000 byte UDP packets to 
the MobiSteer node at a constant rate of 300 packets/sec. The ad hoc mode is chosen 
instead of infrastructure mode so that the MobiSteer node can receive packets from 
any AP rather than only the specific AP it is associated to. This enables us to exclusively
study the beam steering part of our algorithm. 

If and when the MobiSteer node receives any packet it records the tuple 
<location,timestamp, AP, channel, datarate, beam, SNR>. 
The default auto-rate algorithm in the card driver is used for rate adaptation. 

The data collection in the controlled experiments is done fairly conservatively 
to eliminate any source of error. In order to eliminate any possibility of missing packets 
due to beam steering delays (which is already negligible), only fixed beams are used for 
each drive and beams are switched only between drives. So a set of 9 drives on the same path 
gives us data on each of the 8 directional beam plus the omnidirectional beam. 

Each drive is done in a very slow speed (about 10 miles/hour). We have done 8 such sets 
of drives on different days and times in order to analyze the variability of the data. 
Recall that we are using one channel as all our deployed APs are in the same channel.

[Traceset] sunysb/mobisteer/kismet (v. 2007-06-30)

top

the initial version

@MISC{sunysb-mobisteer-kismet-2007-06-30,
  author = {Vishnu Navda and Anand Prabhu Subramanian and Kannan Dhanasekaran and Andreas Timm-Giel and Samir R. Das},
  title = {{CRAWDAD} trace set sunysb/mobisteer/kismet (v. 2007-06-30)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/sunysb/mobisteer/kismet},
  month = jun,  
  year = 2007
}
					

This set of log files includes data traces that were collected from a
moving car equipped with an electronically steerable directinal antenna. 
We drove the car in two different environments in Stony Brook University 
campus - Apartment Complex and Parking lot.

We conducted our experiments to evaluate our beam steering algorithm. 
We use two specific controlled scenarios where we deploy our own APs.  

(a) a large empty parking lot in Stony Brook University campus without any neighboring 
buildings and large trees -- offering a virtually multipath-free environment with little, 
if any, external interference, 
(b) the graduate students' apartment complex in the same campus -- offering diametrically 
opposite environment, rich in both multipath and external 802.11 traffic.  

We use only one AP in the parking lot. It has been hard to use more than one AP gainfully 
in such a "clean" environ ment! However, we use five APs in the apartment complex.  
Here, the APs are carefully located so that at each point on our driving route, typically 
two APs are always heard and all points on the driving route are covered by at least one AP. 
This controlled set of experiments demonstrates the beam steering advantage by doing actual 
measurements of link-layer data transfer rate between the MobiSteer node and the APs. 
The APs are run on the same channel. Using just one channel in the experiments removes 
the channel variable from our experiments and lets us concentrate on only the beam steering 
aspect.

[Trace] sunysb/mobisteer/kismet/apt (v. 2007-06-30)

top

the initial version

@MISC{sunysb-mobisteer-kismet-apt-2007-06-30,
  author = {Vishnu Navda and Anand Prabhu Subramanian and Kannan Dhanasekaran and Andreas Timm-Giel and Samir R. Das},
  title = {{CRAWDAD} trace sunysb/mobisteer/kismet/apt (v. 2007-06-30)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/sunysb/mobisteer/kismet/apt},
  month = jun,  
  year = 2007
}
					

This set of log files includes data traces that were collected from a
moving car equipped with an electronically steerable directional antenna.
We drove the car in an apartment complex in Stony Brook University campus.

This set of log files includes data traces that were collected from a
moving car equipped with electronically steerable Phocus Array Antenna
from Fidelity Comtech. We drove the car in an apartment complex in 
Stony Brook University campus.  We had setup up 5 Accesspoints (802.11g) 
alongside the driving route. The APs were configured to send CBR traffic 
at 300 packets/second to the mobile node. Each file corresponds to the data 
collected at the mobile node using kismet tool during one complete drive 
along the route using one single beam pattern. We collected the data for 
9 beam patterns (given below) by driving 9 times along the same route, 
each time with a different beam pattern. We have included the data collected 
on 12 different days.

Beam-0 - Omni

Directional beams (45deg beam width and 15dBi gain):
Beam-1 - 0deg
Beam-3 - 45deg
Beam-5 - 90deg
Beam-7 - 135deg
Beam-9 - 180deg
Beam-11 - 225deg
Beam-13 - 270deg
Beam-15 - 315deg

Data format is as follows:
    <gps-point ap=apt-ap1 time-sec=1163634436 time-usec=142102 lat=40.544773 lon=-73.066154 type=2 subtype=0 signal=10 noise=0 beamIndex=0 channel=6 datarate=110/>

Datarate field is interpreted as follows
10 -- 1.0 Mbps
20 -- 2.0 Mbps
50 -- 5.5 Mbps
110 -- 11.0 Mbps
60 -- 6.0 Mbps
120 -- 12.0 Mbps
180 -- 18.0 Mbps
360 -- 36.0 Mbps
90 -- 9.0 Mbps
240 -- 24.0 Mbps
480 -- 48.0 Mbps
540 -- 54.0 Mbps

[Trace] sunysb/mobisteer/kismet/parking-lot (v. 2007-06-30)

top

the initial version

@MISC{sunysb-mobisteer-kismet-parking-lot-2007-06-30,
  author = {Vishnu Navda and Anand Prabhu Subramanian and Kannan Dhanasekaran and Andreas Timm-Giel and Samir R. Das},
  title = {{CRAWDAD} trace sunysb/mobisteer/kismet/parking-lot (v. 2007-06-30)}, 
  howpublished = {Downloaded from http://crawdad.cs.dartmouth.edu/sunysb/mobisteer/kismet/parking-lot},
  month = jun,  
  year = 2007
}
					

This set of log files includes data traces that were collected from a
moving car equipped with an electronically steerable directional antenna.
We drove the car in a parking lot in Stony Brook University campus.

This set of log files includes data traces that were collected from a
moving car equipped with electronically steerable Phocus Array Antenna
from Fidelity Comtech. We drove the car in a parking lot in Stony Brook 
University campus. We only had one AP (802.11b) covering the entire area. 
The APs were configured to send CBR traffic at 300 packets/second to 
the mobile node. Each file corresponds to the data collected at the mobile 
node using kismet tool during one complete drive along the route using 
one single beam pattern. We collected the data for 9 beam patterns (given below) 
by driving 9 times along the same route, each time with a different beam 
pattern. 

Beam-0 - Omni

Directional beams (45deg beam width and 15dBi gain):
Beam-1 - 0deg
Beam-3 - 45deg
Beam-5 - 90deg
Beam-7 - 135deg
Beam-9 - 180deg
Beam-11 - 225deg
Beam-13 - 270deg
Beam-15 - 315deg

Data format is as follows:
    <gps-point ap=apt-ap1 time-sec=1163634436 time-usec=142102 lat=40.544773 lon=-73.066154 type=2 subtype=0 signal=10 noise=0 beamIndex=0 channel=6 datarate=110/>

Datarate field is interpreted as follows
10 -- 1.0 Mbps
20 -- 2.0 Mbps
50 -- 5.5 Mbps
110 -- 11.0 Mbps
60 -- 6.0 Mbps
120 -- 12.0 Mbps
180 -- 18.0 Mbps
360 -- 36.0 Mbps
90 -- 9.0 Mbps
240 -- 24.0 Mbps
480 -- 48.0 Mbps
540 -- 54.0 Mbps

[Author] Vishnu Navda

top

[Author] Anand Prabhu Subramanian

top

[Author] Kannan Dhanasekaran

top

[Author] Andreas Timm-Giel

top

[Author] Samir R. Das

top

[Paper] navda-mobisteer

top

In this work, we investigate the use of directional antennas and beam steering 
techniques to improve performance of 802.11 links in the context of 
communication between a moving vehicle and roadside APs. To this end, we 
develop a framework called MobiSteer that provides practical approaches to 
perform beam steering. MobiSteer can operate in two modes - cached mode - where 
it uses prior radio survey data collected during \idle" drives, and online 
mode, where it uses probing. The goal is to select the best AP and beam 
combination at each point along the drive given the available information, so 
that the throughput can be maximized. For the cached mode, an optimal algorithm 
for AP and beam selection is developed that factors in all overheads. We 
provide extensive experimental results using a commercially available eight 
element phased-array antenna. In the experiments, we use controlled scenarios 
with our own APs, in two di erent multipath environments, as well as in situ 
scenarios, where we use APs already deployed in an urban region - to 
demonstrate the performance advantage of using MobiSteer over using an 
equivalent omni-directional antenna. We show that MobiSteer improves the 
connectivity duration as well as PHY-layer data rate due to better SNR 
provisioning. In particular, MobiSteer improves the throughput in the 
controlled experiments by a factor of 2 - 4. In in situ experiments, it 
improves the connectivity duration by more than a factor of 2 and average SNR 
by about 15 dB.