Skip to content

Mobile Communications Group @ UIB

  • Increase font size
  • Decrease font size
  • Default font size
Home Members Jaume Ramis

AM3DIO: Adaptive Multicarrier-Multiantenna-Multiuser networks based on Distributed inference and Optimisation

Research Area: Communications and Networking AM3DIO: Adaptive Multicarrier-Multiantenna-Multiuser networks based on Distributed inference and Optimisation
Status:  
Project leaders: Collaborators:
Proposed start date: 2012-01-01 Proposed end date: 2015-12-31
Description:

Nowadays there seems to be a large consensus on several key technological pillars that will surely form the basis of all 4G proposals. Among others, concepts like multicarrier multiantenna-based physical layer (MIMO-OFDMA), relay-assisted infrastructure and overlay cellular coverage via macrocells and femtocells, will serve 4G systems to offer unprecedented data rates and coverage while satisfying stringent quality-of-service (QoS) requirements. Extending the current use of MIMO (multiplexing/diversity), next generation systems will rely on MIMO processing to exploit the users' spatial separation (Multiuser MIMO) and the possibility of mobile terminals simultaneously linked to several base stations or relays (virtual MIMO). The presence of relays in a cellular network, jointly with an overlay structure, should help improving cell coverage while providing users with higher throughputs. Given a target geographical area, all the aforementioned techniques have as common goal the improvement of the spectral efficiency measured in b/s/Hz/m2.

Project AM3DIO will focus on the study of adaptive schemes in 4G systems equipped with these mechanisms (OFDMA, advanced MIMO processing, cooperative network elements). One of the key problems that remains largely open in 4G networks is how to efficiently manage the multitude of degrees of freedom offered by these new techniques. This issue becomes even more involved when the dynamic character of the environment (channel variations, different user load) must be accounted for. The main goal of this project is the design of schemes that would allow an AM3DIO device (terminal, base station or relay) to adjust its parameters in light of the information provided by the network and that gathered from its surrounding environment with the goal of optimising some performance measure (e.g. capacity, fairness). It is envisaged that this objective will be fulfilled in two steps:

  1. 1)Definition of a unified physical layer (uPHY) framework that is able to offer a high level of granularity of the system resources in space, time and frequency.
  2. 2)A radio resource allocation (RRA) unit tightly integrated with the defined uPHY that optimally distributes the resources among the different users in an adaptive fashion in response to environment changes.

Where possible, the adaptation process should rely on local information in order to minimize signalling overhead while also improving the overall system response in front of rapid changes. It is envisaged that such adaptation and allocation mechanism will be heavily sustained by two powerful mathematical tools: (convex) optimisation and (machine) learning. Optimisation will primarily be used for many of the problems usually encountered in cross-layer resource allocation such as subcarrier, time slot and power allocation, adaptive modulation and coding (AMC) and scheduling. Fundamental (e.g. capacity) or pragmatic (e.g. error rate, QoS) measures can be used to guide the optimisation procedure, which will often require of a cross-layer approach since it will involve mechanisms situated in different layers. Sometimes, a closed-form solution of the problem to be solved may prove elusive, may have some information missing or may be too computationally complex to be solved. In such cases, gathered information (either local or network-wise) may become useful towards the development of adaptive strategies that pursue the same objective as the original optimisation problem. In this case, the derivation of learning strategies based on information inferred in a distributed manner from the network may prove instrumental to the development of agile adaptive strategies able to operate with incomplete information.