Cognitive Collaborative Radio Networks

During the past decades we have witnessed the explosive emergence of wireless technologies and standards, covering different ranges and different licensed and unlicensed spectral bands. This has triggered the appearance of a huge amount of mobile/wireless devices, many of them coexisting in the same environment. Unfortunately, the wireless radio spectrum is a scarce resource, and the available wireless bandwidth does not scale with the needed wireless bandwidth. Certain parts of the wireless radio spectrum, in particular the license-free ISM bands, are overcrowded, while other parts of the spectrum, mostly licensed bands, may be significantly underutilized. As such, there is a need to introduce more advanced techniques to access and share the wireless medium, either to improve the coordination within a given spectral band (horizontal spectrum sharing), or to explore the possibilities of opportunistically using unused spectrum in underutilized (licensed) spectral bands (vertical spectrum sharing). Cognitive radio networking solutions seek to optimize this sub-optimal use of radio spectrum that exists today, by autonomously reconfiguring radio transmission parameters and network configuration settings based on the environment in which the radio devices are operating.

Within this research track, members of IDLab are looking at a novel collaborative cognitive radio approach which is positioned on the intersection between multi-agent learning and software defined radios (SDR).

This research track has two complementary goals:

GOAL 1: Realizing true real-time SDR.

The concept of SDR is very encouraging for the development of state-of-the-art physical layer (PHY) functionality, because software programming allows much faster development cycles. SDR development has so far mostly been limited to non real-time physical layer development, as software implementations do not always offer the fast execution times that are required for true networking experimentation (for example requiring fast acknowledgment of MAC frames within a few microseconds). There is a recent trend to code more and more transceiver functionality on hardware, trading software flexibility for faster execution times, at the cost of higher design time efforts. The IDLab team targets real-time SDR capabilities through hardware/software co-design enabling the optimization of the functionality and performance of transceiver processing units through hardware acceleration on FPGA for computational-intensive processes; time-critical and less computation-intensive processes could run on an embedded microcontroller, non time-critical processes can run on a host PC
Real-time SDR should support the following functionality:

• Low-latency operation allowing very fast MAC response times, in the order of microseconds

• High throughput through hardware/software co-design, enabling operation in mmWave bands and MIMO beamforming

• Real-time end-to-end networking, not just supporting physical layer, but also integrating MAC and higher layer network functionality

• Fully embedded system design no more requiring a host PC for running and controlling SDR

• Open, (over the air) reconfigurable/reprogrammable HW/SW architectures allowing fast prototyping, and runtime configuration and orchestration of transceiver chains and higher layer network functionality.

• Customized QoS per traffic class through network slicing (= virtualization of the physical network infrastructure into independent, customized networks, each applying the most appropriate network protocols) and radio slicing (virtualization of a single physical radio into multiple independent radios, each applying the most appropriate modulation and coding scheme and resource sharing mechanisms).

GOAL 2: The realization of a multi-agent learning approach for cognitive radios

Relying on a real-time SDR component that can dynamically tune the entire protocol stack (from PHY spectrum to higher networking layers), we aim to build an intelligence layer that enables an environment where every participating node (i.e., SDR) can collaborate to optimize the spectrum usage.

Each node in this context can be modeled as an agent in a multi-agent learning environment. We focus on building specific multi-agent learning algorithms that guarantee decision stability and are able to converge in a minimum amount of time. For this, we  investigate the dynamic spectrum access problem within cognitive radios. More specifically, we target a multi-agent reinforcement learning system which supports both complete decentralized scenarios (no communication between agents) and scenarios in which the communication overhead is existing but limited. The multi-agent platform should be capable of supporting the following functionality:

• Dynamic environment: with new nodes joining and leaving the network, learning should be stable and scalable.

• High dimensionality: both at the MAC and PHY layer a plethora of parameters need to be configured. We go beyond simple power allocation and channel allocation and tune the full spectrum of parameters.

• Limited to no communication: as it is embedded in a communication environment with scarce spectrum and sometimes different technologies, nodes can often not communicate with each other, or very little.

• Support for noise and uncertainty: as the wireless medium is in essence an unreliable communication channel there can be both noise and uncertainty about the sensed information.

• Support for incumbent technologies: in the same spectrum, not all  nodes can be learning agents. The learning framework should also support non-rational or non-utilitarian agents.

Staff

Steven Latré & Ingrid Moerman

Researchers

Patrick Bosch, Bart Braem, Miguel Camelo, Bart Spinnewyn, Ruben Mennes
Spilios Giannoulis, Irfan Jabandzic, Xianjun Jiao, Tarik Kazaz, Merima Kulin, Wei Liu, Vasilis Maglogiannis, Felipe Augusto Pereira de Figueiredo

Projects

• FP7 FLEX: FIRE LTE testbeds for open experimentation
• H2020 WiSHFUL: Wireless Software and Hardware platforms for Flexible and Unified radio and network controL
• H2020 eWINE: elastic Wireless Networking Experimentation
• H2020 Flex5GWare: Flexible and efficient hardware/software platforms for 5G network elements and devices
• H2020 ORCA: Orchestration and Reconfiguration Control Architecture
• MoniCow: More Efficient Cattle Monitoring Through an Advanced Data System
• FWO/SBO SAMURAI: Software Architecture and Modules for Unified RAdIo control
• FWO/SBO SMILE-IT
• GOA Disposable and biodegradable wireless networks for extreme conditions

Key publications

• Miguel Camelo, Jeroen Famaey, Steven Latré, “A Scalable Parallel Q-Learning Algorithm for Resource Constrained Decentralized Computing Environments”, Machine Learning in HPC Environments 2016
• Kazaz, T., Van Praet, C., Kulin, M., Willemen, P., & Moerman, I. (2016). Hardware accelerated SDR platform for adaptive air interfaces. ETSI Workshop on Future Radio Technologies : Air Interfaces (pp. 1–10). Presented at the ETSI Workshop on Future Radio Technologies : Air Interfaces.
• Wei Liu, Daan Pareit, Eli De Poorter and Ingrid Moerman. Advanced spec- trum sensing with parallel processing based on software-defined radio. EURASIP Journal on Wireless Communications and Networking, 2013.1 (2013): 1-15.
• Wei Liu, Mikolaj Chwalisz, Carolina Fortuna, Eli De Poorter, Jan Hauer, Daan Pareit, Lieven Hollevoet and Ingrid Moerman Heterogeneous spec- trum sensing: challenges and methodologies. Published in EURASIP Jour- nal on Wireless Communications and Networking, 2015.1 (2015): 1-15
• Wei Liu, Stratos Keranidis, Michael Mehari, JonoVanhie-VanGerwen,Ste- fan Bouckaert, Opher Yaron and Ingrid Moerman. Various detection tech- niques and platforms for monitoring interference condition in a wireless testbed. Published in Lecture Notes in Computer Science, 7586, p.43-60, 2013.