PoQEMON : Platform for Quality Evaluation of Mobile Networks
The PoQEMoN project aims to develop the reference system in France and Europe for measuring the perceived quality of mobile networks and services. This project was supported by the Systematic competitiveness cluster, the state through the FUI (BPI France and the DGE) as well as by the Ile-de-France region.
The 9 partners (Orange, Altran, Next Interactive Media, GenyMobile, Data Publica, HEC, INRIA-Orpailleur and UPEC-Lissi) completed the project which ended on July 1, 2016 and all the results were to be presented to funding agencies as well as to different actors in the targeted ecosystem.
ip-label, as project leader, ensured the coordination of the partners, the progress of the works and the relations with the public authorities. The project management activities materialized through 12 steering committee meetings.
The results obtained
Metrology and Test Infrastructure
ip-label and GenyMobile worked together to define code libraries for developing metrology tools. GenyMobile has brought its expertise in the discovery of low-level APIs on Android system, allowing to capture metrics at the TCP stack level. ip-label completed the development of the Metrology Framework. ip-label has also deployed a network of test servers in Europe, Asia and North America to deliver a global measurement tool.
This network has been deployed on the Amazon cloud on servers with large network traffic capabilities. The servers were qualified by ip-label and GenyMobile by performing automated test batteries. A dynamic routing system has been set up to test the server closest to the user.
Mobile QoE Video Measurement Application
The partners worked on an application dedicated to the study of QoE measurement and QoS / QoE correlations. To do this, the chosen usage was mobile streaming video. Thus, the partners relied on the Metrology Framework to define the operation of the application. Altran coordinated this SP to develop an application with two parts:
- a questionnaire to be completed by the user to determine his profile and collect the assessment of the quality of viewing,
- a YouTube feeds viewer to collect network quality information at the device.
The structuring of the data has also been defined in order to automate the collection of the results in a database and then to carry out the studies. & nbsp; HEC participated in the definition of the test campaigns. Once the specifications were defined, she did a few tests but did a job sharing with the UPEC: while the UPEC was dedicated to conducting intensive tests on the video measurement application, HEC concentrated its efforts on the crowdsourcing app (01net.com SpeedTest).
QoE modeling and QoS / QoE correlation
This part of the project has worked on QoE modeling based on the results of previous work. For this, evaluation campaigns were set up by UPEC and Orange. A total of 2,100 evaluations were conducted by 218 users, which is a very important database for this type of test. In order to structure the data collected, the UPEC proposed a formal model. This database has made it possible to refine the various statistical analyzes that relate to the interaction between the various factors that affect the perception of the user in order to draw discriminant characteristics from them and thus to select the best factors impacting users based on their contexts. These analyzes allowed the UPEC to propose a real user perception prediction model based on a selected factor approach, multi-modular, adaptive and self-regulating using artificial intelligence tools. Finally, the UPEC started a campaign of experimentation and validation of the proposed model by integrating it into a video player implementing the MPEG-Dash standard.
Orange has been particularly interested in the correlation between quality of QoS service, which can be measured by automated tools, and the quality of user experience
