The two main research axes of the Adopnet team are :
Radio-Network Control
The radio access network is no longer a set of base stations, each one working independently from each other, but a group of radio units controlled by a
central unit through a various number of Distributed Units possibly virtualized with the advent of Open Radio Access Networks (O-RAN). Furthermore, most radio access networks mixe several frequency bands, which induces cells of different sizes and thus a multi-layer network. and multi-antenna techniques such as Multi-Input Multiple-Output (MIMO) transmissions are now generalised. They can provide high bit rate but they require intensive processing and high-capacity links between the radio units and the distributed units (fronthaul links). In exceptional circumstances including natural disaster and low-density environment, the fronthaul links are based on radio transmission. Furthermore, a high degree of reliability is required, possibly at the expense of lower capacity.
Our objective is to contribute to the defi nition of new radio-access architectures and associated control procedures that are able to adapt to the varying load conditions regarding both the time dimension, the space and the type of services.
This axis includes studies on
• Radio Resource allocation (scheduling with service di erentiation, power-control, MIMO modes, energy saving),
• Terminal-access point association in a multi-layer context,
• Hybrid networks that combine device-to-device and device-to-network transmissions or radio-based fronthaul,
• Functional split between the radio units and the distributed unit for di fferent fronthaul types and the related optimizations.
Edge-Network Control
Network Function Virtualization (NFV) is a strong trend in networks. It is adopted for example for all Network Functions (NF) of a 5G network. A
service is defi ned by a composition of elementary functions, called VNF (Virtual Network Functions), which can be deployed at diff erent locations on the network,
potentially operated by di fferent actors. It thus allows the emergence of virtualized or non-virtualized service providers and their composition.
In addition, network slicing allows to de fine several virtual networks dedicated to speci c use cases. The slices are implemented with di fferent levels of isolation on the same physical infrastructure, which is potentially complex and operated by di fferent actors. Depending on the type of use case they address, slices must respect a set of properties that can range from properties of availability, quality of service (latency, jitter, …) to properties related to energy consumption or security. Ful lling constraints on these properties is challenging, especially for use-cases implying dynamic “edge-to-edge” communications.
Therefore, enabling slicing requires mechanisms to ensure their dynamic adaptation to the network conditions involving self-con guration, monitoring, analysis and planning. However, the objective is not to design fully self-organized networks but to put these mechanisms at the service of the implementation of a global strategy. An important issue is to determine when auto-adaptation actions should be performed in a distributed way or require a centralized approach and to which extent hybrid approaches can be used.
One question is how to defi ne the slices, where to locate the involved virtual network fuctions in order to provide the expected quality of the global network service especially in term of latency and throughput while minimizing the energy consumption of the system and ensuring a minimum resiliency against failures. We will consider automatic scaling and placement of these functions as part of the solution to dynamically adapt the slicing to changes in the initial conditions.
