PhD Day 2017 @ CITI-Lab

Florent de Dinechin

Academic careers

30 mins

Amphi Chappe

Introduction

Details about the PhD Day and introduction of the invited talks.

15 mins

Telecom building hall

Poster and demo session 1

45 mins

Telecom bulding hall

Antoine Fraboulet

La transformation de la recherche de laboratoire à la R&D dans une startup.

30 mins

Amphi Chappe

Plenary session 1

60 mins

Amphi Chappe

Poster and demo session 2

45 mins

Telecom building hall

Jean-Christophe Sibel

To be announced

30 mins

Amphi Chappe

Plenary session 2

30 mins

Amphi Chappe

Conclusion & Cocktail

Conclusion on CITI PhD-Day 2017 and a very welcomed cocktail.

60 mins

Hall

Guillaume Bono

(Reinforcement) Learning to solve Rich Vehicle Routing Problems

Vehicle Routing Problems (VRP) are a very challenging class of combinatorial problems where a fleet of vehicles must deliver goods to a set of locations. Solutions aim at optimizing both the assignment and the ordering of the deliveries under constraints. There exist many refinements of the problem, considering stochastic travel times, dynamic clients, transfer capabilities, limited energetic autonomy... which can be combined under the Rich Vehicle Routing Problem (RVRP) class. Our approach consists in combining a decentralized structure inspired by multi-agent systems to split the assignment and routing tasks between different hierarchical levels, and a reinforcement learning approach based on Deep Q-Networks to automatically learn features and approximate value functions used to guide decisions.

45 min

Telecom building hall

Team

Chroma

Tristan Delizy

Towards Dynamic Scratch Pad Memory Management for Embedded Systems

New memory technologies have been emerging during last years under the name of non volatile RAM. These different technologies promising memory faster and denser than DRAM have drawn attention in the high performance computing domain but still struggle to overcome limitations in these applications. In the same time, along with the growth of IoT domain these memories has been considered to be integrated directly inside Systems on Chips targeting Embedded Systems applications. In such systems, running at relatively low frequencies in front of HPC architectures, NVRAM can be integrated directly next to the CPU and allow to exploit these technology density to build inSoCs large Scratch Pads Memories. These architectural implications leads to new programming models like normally off execution, focusing on time and energy efficiency of such systems. But the arrival of inSoC heterogeneous memory hierarchy exposed to the software brings new difficulties for development of embedded software and harden the reuse of legacy code. We propose to look at runtime dynamic memory management new opportunities brought by these memory innovations considering the domain of low power devices for IoT.

45 mins

Telecom building hall

Team

Socrate

Andrea Bocco

Hardware support for UNUM floating-point arithmetic

UNUM (Universal NUMber) is a new floating point format designed to replace the standard one (defined in the IEEE 754 standard). The aim of this work is:

• Design an hardware implementation of UNUM arithmetic.
• Assess its cost compared to a conventional IEEE 754 FPU.
• Propose a co-processor architecture to embed UNUM operations in a working computation core.

45 mins

Telecom building hall

Team

Socrate

Diane Duchemin

Distributed coding for Ultra Reliable Low Latency massive IoT networks

In the context of massive Cellular IoT Networks, we tackle the problem of designing an efficient Medium Access Control (MAC) for multiple Low Power Wide Area (LPWA) and Ultra Reliable Low Latency (URLL) communications. We first highlight the constraints of those types of networks before proposing the model of a solution, based on Compressive Sensing (CS) theory and Non Orthogonal Multiple Access (NOMA) scheme.

45 mins

Telecom building hall

Team

Socrate

Victor Morel

Protecting personal data in the smart city

The exponential growth of IoT devices is leading to a extensive collection of data in ubiquitous environments, notably personal data. This collection can threaten our privacy if the greatest care is not taken. We will examine in this poster what are the risks stemming from personal data collection, applied to a smart city use case, and present potential solutions to avoid a "Orwellian nightmare" scenario. These solutions (such as the visibility of devices and their data practices, as well as the transmisson of privacy policies), although presented very generically, could be implemented technically in existing protocols.

45 mins

Telecom building hall

Team

Privatics

Ahmed Boubrima

Wireless Sensor Networks for Air Pollution Monitoring (demo)

45 mins

Telecom building hall

Team

Agora

Alexis Duque

LightIoT: Bi-directional Visible Light Communications for IoT Devices (demo)

45 mins

Telecom building hall

Team

Agora

CorteXlab

CorteXlab (visit)

45 mins

Telecom building hall

Team

Socrate

Ahmed Boubrima

Wireless Sensor Networks for Air Pollution Monitoring

Air pollution affects human health dramatically. According to the World Health Organization (WHO), the exposure to air pollution is accountable to millions of casualties every year. As a consequence, the reduction of pollutant emissions is at the heart of many sustainable development efforts. Current air quality measuring systems are massive, inflexible and expensive. An alternative solution would be to use wireless sensor networks (WSN). Although some WSN-based air quality monitoring systems are already operating, the deployment issue of these tiny nodes while taking into account the precision of the resulting network has not yet been investigated. In this talk, we present novel optimization models for WSN deployment while considering the application case of air pollution monitoring and the communication characteristics of sensor nodes.

15 mins

Amphi Chappe

Team

Agora

Alexis Duque

SeedLight: hardening LED-to-Camera communication with Random Linear Coding

During this talk, I will introduce SeedLight, a coding scheme designed to face the inherent packet losses in line-of-sight LED-to-Camera communication and enhance such communication system goodput. SeedLight leverages random linear coding to provide an efficient redundancy mechanism that works even on PHY-SDU of tens of bits. The key idea of SeedLight is to reduce the code overhead by replacing the usual coding coefficients by a seed. Since this work addresses IoT devices with low computational resources, SeedLight encoding algorithm complexity remains low. The experimental evaluation shows that the achievable goodput can be up to 2.5kbps, while the gain compared to a trivial retransmissions scheme is up to 100%.

15 mins

Amphi Chappe

Team

Agora

David Kibloff

Broadcast Channels with Covert Messages

In this talk, the maximum amount of information that can be covertly transmitted over the two-user discrete memoryless broadcast channels is investigated. In particular, given a broadcast code to reliably transmit information to the two users, the maximum number of messages that can be covertly embedded in the broadcast code and reliably transmitted to the intended user is established. As in the point-to-point case, the size of the covert codebook is shown to scale with the square root of n, where n is the blocklength.

15 mins

Amphi Chappe

Team

Socrate

Yuqi Mo

Optimization of IoT Networks based on Ultra-Narrow-Band

UNB (Ultra Narrow Band) is one of the LPWAN technologies. Dedicated to small packets, low energy consumption and long-distance transmission, UNB has been deployed in the SigFox's networks around the world. The natural channel access scheme of UNB is R-FTMA : Random Frequency and Time Multiple Access, where transmissions are realized randomly, both in time and frequency. This channel access scheme saves the resource for reservation, but it also induces some simultaneous transmissions in the same frequency area, leading to interference. This talk gives some insights to optimize the capacity of UNB based networks, by considering its spectral interference. For example the use of replication mechanism, the interference cancellation technology, and the joint signal detection by multiple base stations.

15 mins

Amphi Chappe

Team

Socrate

Hassan Kallam

Topology aided Multi-User Interference Management in Wireless Networks

Topological interference management (TIM) problem, refers to the study of the degrees of freedom (DoF) of wireless communication networks subject to interference. The particularity of TIM is that, unlike the interference channel (IC) models, a TIM model does not rely on any information on the channel states (CSI), but rather only the network topology. The network topology is a connectivity graph that carries information of desired links and interference links, in which a distinction between weak (non-significant) and significant channels are made. The subject of this thesis is the TIM problem for downlink cellular networks under unplanned and chaotic deployment, for example, the case of disaster-relief or military type deployments or small-cells. In this poster, we consider the interference topology problem, i.e., classifying the interference links as significant or weak interference links.

45 mins

Telecom building hall

Team

Socrate

Nizar Khalfet

Simultaneous Information and Energy Transmission in the Interference Channels

In this poster, the fundamental limits of simultaneous information and energy transmission in the two-user Gaussian interference channel (G-IC) with and without feedback are fully characterized. More specifically, an achievable and converse region in terms of information and energy transmission rates (in bits per channel use and energy-units per channel use, respectively) are identified. In both cases, with and without feedback, an achievability scheme based on power-splitting, common randomness, rate splitting, superposition coding and Markov backward decoding is presented. Finally, converse regions for both cases are obtained using some of the existing outer bounds for the information transmission rates, as well as a new outer bound for the energy transmission rate.

45 mins

Telecom building hall

Team

Socrate

Jonathan Tournier

Security evaluation of a IoT network

The fast growth of the IoT is leading manufacturers to provide devices as fast as possible pushing the security of these new systems into the background. This problem is well known in classic network system and number of solution can fill the lack of security such as the penetration testing. However, the IoT raises new problems such as the scalability, the systems heterogeneity and the impressive number of protocols (proprietary or not). We want to propose a solution inspired by the penetration testing but suitable for the context of IoT.

45 mins

Telecom building hall

Team

Dynamid

Jean-Baptiste Trystram

P2P distribution of software updates in IoT environments

The number of connected devices around us is growing, and their embedded software is increasing in complexity. Applications need to be managed and updated throughout their life-cycle. Modern software orchestrators and management systems are mostly centralised and expensive to scale to very large systems. We propose a decentralised approach to distribute software updates to Internet of Things (IoT) devices using Peer-to-Peer (P2P) transfers. We seek to improve the efficiency of software update propagation by adapting system behaviour to IoT constraints, such as device heterogeneity, unreliable network connectivity and applications specifics.

45 mins

Telecom building hall

Team

Dynamid

Yohann Uguen

High-level synthesis and arithmetic optimizations

FPGAs are well known for their ability to perform non-standard computations not supported by classical microprocessors. Many libraries of highly customizable application-specific IPs have exploited this capablity. However, using such IPs usually requires handcrafted HDL, hence significant design efforts. High Level Synthesis (HLS) lowers the design effort thanks to the use of C/C++ dialects for programming FPGAs. However, high-level C language becomes a hindrance when one wants to express non-standard computations: this languages was designed for programming microprocessors and carries with it many restrictions due to this paradigm. This is especially true when computing with floating-point, whose data-types and evaluation semantics are defined by the IEEE-754 and C11 standards. If the high-level specification was a computation on the reals, then HLS imposes a very restricted implementation space. This work attempts to bridge FPGA application-specific efficiency and HLS ease of use. It specifically targets the ubiquitous floating-point summa\-tion-reduction pattern. A source-to-source compiler transforms selected floating-point additions into sequences of simpler operators using non-standard arithmetic formats. This improves performance and accuracy for several benchmarks, while keeping the ease of use of a high-level C description.

45 mins

Telecom building hall

Team

Socrate

Régis Rousseau

Optimization of Wireless Power Transfer System

With the advent of IoT, the problem of feeding communicating objects with low energy resources becomes crucial. Minimizing the energy consumption of these objects and collecting different forms of energy like the transfer of power through electromagnetic waves are some solutions found. In this thesis, an overall view of the power transmission system is based on the waveform, propagation channel and collection circuit aspects in order to maximize the amount of energy harvested.

45 mins

Telecom building hall

Team

Socrate

Etienne Villedieu

Storing IoT data in the BitTorrent DHT

Large scale IoT systems can be hard to manage and to be fault-tolerant. One solution may be to distribute data storage and decision making over the differnets nodes. This way, if some elements of the network shuts down or have a non wanted behavior, the other nodes can take the relay. Distributed hash tables can handle some of these problems by providing a distributed network across IoT devices.

45 mins

Telecom building hall

Team

Dynamid

Ben Khalifa Abderahman

Medium Access Control in Cognitive IoT Networks

Cognitive networks are based on self-awareness and context-awareness requirements. Recent works focused on adding cognition in the operation of upper layers, while medium access control in Wireless Networks remains based on static approaches (reduced access to storage and computing resources) Cognitive mechanisms applied in the MAC layer could benefit from Aloha and CSMA approaches, adding scala-bility, reliability and quality of service to IoT MAC solutions. For example, default MAC layer settings could be defined for a certain node density and activity, and the network would learn and adapt when these default conditions change.

45 mins

Telecom building hall

Team

Agora

Gautier Berthou

Running IOT applications while facing frequent power failures (demo)

Systems using energy harvesting as only power source may have high energy requirements, for instance when sending a packet over a radio channel. It is a common approach to design the power source so that the application runs without power failure, RFID readers are a typical example. However this would not work if the power source is scarce or if the energy storage is not large enough to enable the application to run to completion without power failure. We propose Sytare, a lightweight OS that allows the application to run over power failures while achieving progress. Using non-volatile memory, we are able to manage application state and peripherals state so that application code does not need to be aware of power failures.

45 mins

Telecom building hall

Team

Socrate

Othmane Oubejja

Reliable and Reproducible Radio Experiments in FIT/CorteXlab SDR testbed: Initial Findings (demo)

45 mins

Telecom building hall

Team

Socrate

CorteXlab

CorteXlab (visit)

45 mins

Telecom building hall

Team

Socrate

Victor Quintero

Interference Channel with Noisy Feedback

The current and future wireless communications systems are characterized by the deployment of a high number of communication nodes. These systems are more limited by interference than noise given the huge number of nodes and the high reuse of limited resources e.g., the number of available communication frequencies. This talk shows interesting results from the theoretical point of view on the benefits of considering channel-output feedback in wireless communications systems to manage the interference and the impact on these benefits when the feedback is impaired by noise. These benefits are in terms of enlarging the capacity region. This presentation will also highlight the use of deterministic models to analyze Gaussian networks.

15 mins

Amphi Chappe

Team

Socrate

Mihai-Ioan Popescu

Dynamic Target Allocation for Multi-Robot Evolving Tasks

The talk is focused on introducing dynamic multi-robot routing, as a continuous adaptation of the well-known robot task allocation process (MRTA). As robot missions have become reality, a growing number of applications requires teams of mobile robots to autonomously accomplish task groups. Target allocation problems have frequently been treated in contexts such as multi-robot rescue operations, exploration, or patrolling, being often formalized as multi-robot routing problems. Falling in the NP-hard class, TSP-based heuristics have been proposed, as well as auction-based methods, but none of the approaches considered the adaptation of evolving robot missions (which may grow up to a saturation bound). We propose a framework for dynamically adapting in time the robot missions to new discovered targets, by assuming multi-hop communications between the exploring robots and periodic connectivity when integrating targets for the mission robots. As a case of study we consider the problem of multi-agent patrolling.

15 mins

Amphi Chappe

Team

Chroma

Yosra Zguira

Design of IoB-DTN protocol for mobile IoT Application to smart bike sharing systems

Communication is essential to the coordination of public transport systems. Nowadays, cities are facing an increasing number of bikes used by citizens therefore the need of monitoring and managing their traffic becomes crucial. In this talk, I will introduce "IoB-DTN", a protocol based on the Delay/Disruption Tolerant Network (DTN) paradigm adapted for an IoT-like applications running on bike sharing system based sensor network. The performances evaluated of three variants of IoB-DTN using the four proposed buffer management policies will be presented.

15 mins

Amphi Chappe

Team

Agora