Communication Systems and Networks

The research focuses on the physical layer and the MAC layer of modern wired and wireless communications systems. Transmitters, receivers and resource allocation are designed and optimized for OFDM(A) based systems, multi-antenna systems (MIMO, distributed MIMO, relay techniques) and multiuser systems. At the receiver side, both detection and estimation are considered. Iterative or turbo based techniques are receiving a lot of attention. Particular attention is paid to multicell wireless communications and sensor networks.

This researches focuses on ultra wide band (UWB) based localisation or positioning. Methods considered are time of arrival (TOA), time difference of arrival (TDOA) and angle of arrival (AOA). Bounds have been derived to assess the potential of UWB, understand the impact of multipath propagation and investigate the ambiguities. Practical estimators are proposed and their performance is investigated. A practical testbed has been developed and is being upgraded. An accuracy of a few millimeters has been achieved for indoor positioning over distances of about ten meters and with obstacles.

The DSL activity is investigating signal processing solutions, mainly at the physical layer, for improving the performance of communication on the telephone lines (digital subscriber lines). The current focus is on the mitigation of the crosstalk between lines, including pre-/post-cancellation techniques, adaptive crosstalk channel estimation, as well as dynamic spectrum management by means of centralized optimization.

The radio propagation activity aims at the measurement, characterization and modeling of multi-dimensional propagation channels for a range of wireless systems and networks: MIMO, WLAN/WPAN, body area networks, vehicular communication systems, etc. The facilities include a state-of-the art channel MIMO sounder, various network analyzers and antennas, as well a ray-tracing tool.

• evaluation of link budget for satellite systems, deep-space missions, ...
• generation of time series for the design of fade mitigation techniquespropagation modelling into collapsed buildings and harsh environments for radio-localization

We develop new protocols and mechanisms for the global Internet. Our main area of expertise are the network and transport layers of the TCP/IP stack. More specifically, we have proposed extensions to allow routing protocols (both intradomain and interdomain) to better handle link failures and have designed and evaluated various types of traffic engineering techniques. We also address the multihoming problem both from the network viewpoint and the endhost viewpoint. We carry both theoretical research and applied research by developing reference implementations of new protocols. We actively participate in the development of future Internet architecture proposals and contribute to standardisation within the Internet Engineering Task Force.

Image and video compression algorithms are investigated, including for stereo and multi-view contents. Visually pleasant and fluent video streaming or image browsing are implemented by adapting compression and forwarding mechanisms to network and terminal resources. This implies the rate-distortion optimization of image/video packet schedules, but also adaptive switching between multiple versions of the content. For low bandwidth wireless accesses, interactive streaming architectures are investigated to allow the end-user to control the trade-offs involved when reducing the spatial and temporal resolution of the streamed content.

Distributed systems are becoming ever larger and must handle problems of frequent faults, security issues, and management complexity.  We have built large-scale transactional storage systems based on techniques from self-organizing peer-to-peer networks.  We are developing design techniques for these systems based on ideas from complex systems theory.  Future work will focus on building practical programming platforms that provide cloud computing properties such as elasticity and that support large-scale applications using techniques from machine learning. 

Network data rates increase tremendously thanks to capabilities opened up by fiber, 5G networks and multi-hundred-gigabit network equipments. However, this progress towards faster network speeds presents a challenge in achieving low-latency Internet services while lowering energy usage of network equipments and services to meet the decarbonization objectives of a greener it. We prototype software network functions (NFV) like firewalls, intrusion detection systems, or load balancers scaling beyond 100Gbps link speeds. Core techniques involve avoiding inter-core communications, careful low-level considerations, research on multi-core software systems and their relation to high-speed interconnects. On top of software pipeline optimizations, we leverage the novel programmable network infrastructure, such as SmartNICs and P4, to develop new intra-server and inter-server load-balancing techniques and exploit the full potential of modern hardware to maximize datacenter efficiency.