Research Topics

My research is about 4 fields:

  • Wi-Fi networks indoor positioning,
  • Mobility analysis based on heterogeneous data sources,
  • Radio resources allocation optimization in MBMS 3G networks,
  • Network communication enabling smart grids.

Wi-Fi networks indoor positioning

Positioning consists in calculating a mobile device geographical coordinates in order to provide its user with context-aware services. I choose to use Wi-Fi signals to enable positioning. First, this network standard is widely deployed. Second, it allows to merge positioning and communication operation within the same device and standard. On top of that, Wi-Fi provides a high data transfer rate, thus enabling rich multimedia content to be broadcast to its users.

I started working on this topic during my PhD thesis (see the PhD thesis section). I proposed several Wi-Fi-based positioning models. This work was used in a project within a partnership with californian startup Cozybit, leading to a U.S. software patent (patent n°20120244875).

This work also lead to several implementations:

  • A radiotap based software deployed on mini-PC with Atheros Wi-Fi NIC. These computers are used to get RSSI measurements from the surrounding devices,
  • Implementation on OpewWRT, with two different versions:
    • first implementation on Linksys WRT54GL access points (Broadcom Wi-Fi chip with dedicated monitor header named Prism header
    • Second implementation on TP-LINK MR3220v2 (Atheros Wi-Fi chip supporting radiotap)
  • A C++ positioning server embedding the models proposed during my PhD thesis,
  • A J2E application with a flexible architecture making the integration of new models easy.

J2E and OpenWRT softwares are fully built over HTTP and can be used in a state of the art Wi-Fi network.

Tools used in these softwares are the following:

  • C++ with boost library,
  • MIPS cross toolchain, C language for OpenWRT programming,
  • GNU microhttpd, libcurl, libpcap et libpthread libraries for enabling HTTP communication and access points radio channels sniffing,
  • JDBC and PostgreSQL for managing server database and J2E access.

There is also an Android client application for graphical demonstrations.

Mobility analysis

My activity in mobility analysis is supported by several government-funded projects in partnership with public administration and private contractors, both interested in knowing and get graphical indicators about their users' mobility. On this topic, I managed the PGAM project (Plateforme Géographique d'Analyse de la Mobilité standing for mobility analysis geographical platform). This project, in partnership with the startup WebGeoServices, aims at proposing and promote mobility analysis models and SaaS graphical tools.

Data used to model mobility are numerous: polls, government statistics studies, GPS tracks, mobile phone companies data, etc.

From the data, we build, quantify, and qualify flows of people. These can be filtered and displayed. The SaaS platform is developped by WegGeoServices. We also use our own internal software tool, based on WxWidgets and OpenGL to test our models.

My project team is composed of 5 persons.

MBMS radio resources optimization

This work aims at proposing an optimization process for grouping users requesting the same data within multicast groups. Building multicast groups in a 3G network is complex due to heterogenous radio link quality (distance towards node-B, topology and buildings), and to the limited radio resources.

The problem is complicated by a second constraint: cell power consumption. Indeed, power consumption is based on link quality and data rate, and usually requires a tradeoff with quality of service.

Preliminary results are promising and far more performant than usual simple methods. However, it requires to run faster in order to be used in a real 3G network.

Our solution is based on simultaneous use of various transmission modes: FACH for broadcast, DCH for dedicated channels and HS-DSCH for shared channels. We use two metaheuristics: simulated annealing, and tabu search. Tabu search performs better than simulated annealing.

Tools implemented in this work are part based on Matlab for modeling UMTS radio channel, and part based on C++ and boost for the optimization engine.

Smart Grids

This project addresses many recent problems rising from sustainable developpment, renewable energies and electric vehicles (both battery EV and fuel cell EV are addressed). European Union is pushing car manufacturers to increase EV market share, thus raising new questions. Our work starts with a partnership with the startup Freshmile, whose goal is to propose a fully integrated service to help EV customers to increase their ROI.

Basically, EV are used as electric energy storage when the power grid is over-producing, and as a backup when it is overloaded. Many problems arise from this system:

  • Global system knowledge (grid load, EV status, driver's habits, electricity pricing),
  • Fast command response time
  • Efficient optimization
  • Batteries and fuel cells accurate physical modeling
  • Robustness

My part in this work is to propose and prototype the overall networking infrastructure. It starts with the ISO/IEC 15118 standard study, enriched with physical simulation and optimization algorithm for the system control. The prototype is developped in C (EV and energy supply controllers) and Android for a user application used to register EV user habits.


My publications are available in the publications section.