Throughput Prediction on 60 GHz Mobile Devices for High-Bandwidth, Latency-Sensitive Applications

Data from our PAM 2021 publication used for throughput prediction in static and mobile scenarios

If you use this dataset in your publication, please cite us as follows:

@InProceedings{aggarwal:pam2021,
author={Aggarwal, Shivang and Kong, Zhaoning and Ghoshal, Moinak and Hu, Y. Charlie and Koutsonikolas, Dimitrios},
title={Throughput Prediction on 60 GHz Mobile Devices for High-Bandwidth, Latency-Sensitive Applications},
booktitle={Passive and Active Measurement (PAM)},
year={2021}
}

Dataset Description

Each folder contains file(s) totalling 10 hours of throughput data under 3 scenarios:

1. Static (Section 3.2)
2. Mobile (Section 3.3)
3. Applications (Section 3.4)
   a. VR
   b. ABR

Please refer to the corresponding sections in the paper for more details regarding the scenarios.

Each file contains the following data for every 10 ms (each row in the CSV file):

• Android sensor data
  • Positional data (Azimuth, Pitch, Roll)
  • Accelerometer data (along x-, y- and z-axes)
• 60 GHz link data (reported by the wil6210 driver)
  • MCS
  • Tx Sector
  • Link Status
  • Signal Quality Indicator (SQI)
  • RSSI

Hardware Setup

Below is an image of the device setup used to collect this dataset.

As can be seen, the smartphone (ASUS ROG Phone II) is placed in a Google Cardboard VR Headset. There are basically 3 dimensions of motion for the smartphone:

1. Rotation w.r.t. the Azimuth axis (X-axis in figure)
2. Rotation w.r.t. the Pitch axis (Y-axis in figure)
3. Translation motion along the Slide axis

This setup which allows us to simulatenously move the smartphone in all 3 dimensions is enabled by using the Cinetics Lynx 3-Axis Slider. Please take a look at the GIF in the hardware-setup folder for a visual demonstration of how the slider moves (source).

The Dragonframe software allowed us to programmatically control the Lynx setup to perform custom mobility patterns at various speeds.