Estimating Video QoE in 5G Networks Using Machine Learning - Supplementary Material

This is the simulation enviroment and supplementary data as used in paper: Accuracy vs. Cost Trade-off for Machine Learning Based QoE Estimation in 5G Networks

An extension to: Schwarzmann, S., Cassales Marquezan, C., Bosk, M., Liu, H., Trivisonno, R., & Zinner, T. (2019, October). Estimating Video Streaming QoE in the 5G Architecture Using Machine Learning. In Proceedings of the 4th Internet-QoE Workshop on QoE-based Analysis and Management of Data Communication Networks (pp. 7-12). ACM.

Simulation Environment

Requirements

Vagrant - Tested with version 2.2.4 (available at: https://www.vagrantup.com/downloads.html)

Vagrant box setup

In the /vagrant-omnet-simulation-mobility/simulation-environment folder use:

$ vagrant up

And wait until the vagrant box setup is complete.

Using the Omnet++ simulator

To access the box and the simulation environment use the following command while remaining in the /vagrant-omnet-simulation-mobility/simulation-environment folder:

$ vagrant ssh

This will establish an ssh connection to the box. In there you can access the simulation environment.

Running a simulation

In order to run simulation scenarios (as they were run within the project) switch to the Moving1/simulations folder:

$ cd Moving1/simulations

and use the following command to run an exemplary simulation scenario:

$ ./run_sim_campaign.sh -c swimMovementFP_ds0_p2_s50-50 -t 1

This will run a simulation scenario with the config [Config swimMovementFP_ds0_p2_s50-50].

-c argument specifies the scenario's config name. You can find what simulation scenarios are available in the Moving1/simulations/omnetpp.ini file.

-t argument specifies the number of threads that will be used to parallelize the simulation.

Adjusting the movement of UEs

In the simulation we include 3 main types of UE movement - pedestrians (at 1mps), vehicles (at 10mps) and stationary clients. All of these can be adjusted in the Moving1/simulations/omnetpp.ini file. In each "Config" section (eg. [Config swimMovementFP_ds0_p1_s100-0]) the following can be used to set speed of UEs (examples):

• *.ue[0..39].mobility.speed = 1mps Sets movement speed of UEs 0 through 39 to 1 mps (pedestrian)
• *.ue[40..79].mobility.speed = 10mps Sets movement speed of UEs 40 through 79 to 10 mps (vehicle)
• *.ue[*].mobility.speed = 10mps Sets movement speed of all UEs to 10 mps (vehicle)

In order to make UEs stationary the following code can be used:

*.ue[0..55].mobilityType = "LinearMobility"
*.ue[0..55].mobility.speed = 0mps

In this example UEs 0 to 55 will become stationary.

Other Simulation settings

The following table shows the most relevant parameters used to configure the eNodeB in the simulated scenarios:

AMC = Adaptive Modulation and Coding
MAC = Medium access control
RLC = Radio link control

Links

1. Omnet++ - https://omnetpp.org
2. SimuLTE - https://simulte.com
3. INET Framework - https://inet.omnetpp.org
4. SWIM Mobility Model GitHub - https://github.com/ComNets-Bremen/SWIMMobility
5. Original version of the TCP Video Client by Joaquín Navarro - https://github.com/navarrojoaquin/adaptive-video-tcp-omnet
6. TCP Video Client Adaptation which the TCP Video Client used in this work is based on - https://github.com/inet-framework/inet/tree/topic/tcpvideosteaming/src/inet/applications/tcpapp (files TcpVideoStreamCliApp.*)

Supplementary Material

Number of features, feature lists and monitoring points

See the table .

It shows, in the case of same number of features, the corresponding feature lists, monitoring points when alpha is the value that result in the lowest Median absolute error (medAE) while applying Lasso Regression.

All, Moving, Pedestrian, Vehicle, Stationary are different type of datasets in experiments.

Contact

1. Susanna Schwarzmann - [email protected]
2. Marcin Bosk - [email protected]
3. Huiran Liu - [email protected]