Marie Curie IRG 224755
The initial cellular systems deployed in the 1980s and 1990s featured conservative frequency reuse patterns in order to ensure a high signal-to-interference-plus-noise ratio (SINR). This allowed operating the links comfortably with limited signal processing at the expense of having a small number of concurrent links. Altogether, the system spectral efficiency (bits/s/Hz/cell) was low and soon became insufficient to satisfy the explosive growth in demand for wireless communication. With the soaring cost of bandwidth and of real estate, the need for higher efficiencies became dire. Since then, we have witnessed a sustained improvement in system spectral efficiency driven, mostly, by advances in communication theory and by Moore’s law. Specifically, the successive introduction of advanced techniques (forward error correction, power control, link adaptation, incremental redundancy, etc) and the massive increases in processing power have enabled:
A progressive rise in link spectral efficiency, which in emerging systems (3GPP Long-Term Evolution, IEEE 802.16 WiMAX) is already approaching capacity.
Operation at diminishing SINRs, enabling ever more aggressive frequency reuse patterns. In fact, the aforementioned systems are reaching the point of universal frequency reuse and are thereby limited first and foremost by their own interference.With link efficiencies approaching capacity and with frequency reuse reaching universality, this marks the end of the road for the approach followed thus far to improve the system spectral efficiency.
In recent years, the introduction of multiple-input multiple-output (MIMO) techniques has provided powerful new means for enhancing the performance of wireless systems. MIMO techniques enable spatial frequency reuse within each cell, but still subject to the high levels of interference from other cells. It is becoming increasingly clear that major new improvements in spectral efficiency will have to entail addressing such intercell interference. Such is, precisely, the premise of NetMIMO.
In the traditional modus operandi of cellular systems, a user is assigned to a cell site on the basis of certain criteria (e.g., signal strength) and it then communicates with that cell site while causing interference to all other sites in the system. The key tenet of NetMIMO is that, in the uplink specifically, intercell interference is merely a superposition of signals that were intended for other cell sites, i.e., signals that happen to have been collected at the wrong place. If these signals could be properly classified and routed, they would in fact cease to be interference and become useful in the detection of the data they bear. (A dual observation can be made about the downlink.) This insight naturally leads to the conclusion that, ultimately, the goal should be to serve all users through all the sites within their range of influence. While challenging, this is theoretically possible by virtue of the fact that the cell sites are connected by a powerful backbone network. This ambitious idea leverages the almost unlimited bandwidth available in optical-fiber wireline networks to transcend the burden of wireless intercell interference. Operationally, this can be interpreted as a form of MIMO that, through the backbone network, spans the entire system as opposed to only each cell separately. Hence the term NetMIMO. With NetMIMO, in fact, the notion of a cell gets blurred once users are no longer assigned to specific sites. Ultimately, there is a network of sites serving a population of users. While this is a conceptually simple proposition, it poses numerous hurdles and challenges that this project aims to resolve.
Objective 1: Establish techniques for NetMIMO uplink detection.
Both linear and nonlinear MIMO detectors need to be considered and classified, and the corresponding tradeoffs between complexity and performance need to be established. Emphasis must be placed on the fact that the signals being received may exhibit large disparities in average strength and in distribution between cell sites. (The signal transmitted by a certain user, for instance, may be received at high average power levels and subject to only modest fading at nearby sites while simultaneously being received at lower average power levels and with more pronounced fading at distant sites.) This situation is markedly different from typical point-to-point MIMO scenarios, where such discrepancies are minor or nonexistent.
Objective 2: Establish techniques for NetMIMO downlink detection.
This problem is the dual of the one addressed under Objective 1. Again, linear and nonlinear MIMO techniques may be considered although, in this case, linear schemes are strongly favored. The issue of average signal strength and fading distribution disparities, identified under Objective 1, remains equally problematic for the downlink. Recall also that, with respect to typical single-cell MIMO setups with a unique power constraint, the NetMIMO setup entails a separate power constraint per site. This renders the NetMIMO downlink significantly more complex than the uplink and it precludes the direct applicability of existing single-cell solutions.
Objective 3: Evaluate the performance of uplink and downlink techniques under various channel estimation policies.
This is a key challenge given that all the preliminary evidence available at this point [1-6] relies entirely on the assumption that the fading states are known instantaneously and perfectly by all the sites in the system. The NetMIMO proposition will be credible only once this assumption has been overcome.
Since the fading distribution may in general not be known a priori, linear channel estimators are preferred. The performance of various time-frequency-space interpolators is to be asserted. In particular, joint multidimensional estimators will have to be evaluated and compared with simpler estimators that operate separately on each dimension (time, frequency and space).
Objective 4: Design reference signal structures specifically tailored to the channel estimation needs of NetMIMO.
Robust reference signal structures should be devised to enable the required estimation quality in the face of a wide range of signal strengths, mobility levels, frequency selectivity, and spatial correlation. Adaptivity of the reference signal structure to these various channel features should be contemplated. The reference-signal overhead should be quantified in order to ensure that it does not erase a substantial part of the NetMIMO gains.
Objective 5: Develop calibration mechanisms to warrant channel reciprocity in time-division duplex (TDD) systems.
NetMIMO is particularly well suited for TDD systems, where the short-term reciprocity of the radio channel allows taking full advantage of the uplink-downlink duality. The reciprocity, however, does not extend to the equipment (amplifiers, filters, etc). Closed-loop calibration mechanisms must be in place in order to compensate for the equipment asymmetries in an efficient and reliable manner.
Objective 6: Quantify tolerance to latency and delay jitters.
Demanding levels of performance by the backbone network are required in order for NetMIMO to function, specifically, high speed and very low latency. Both centralized and distributed NetMIMO architectures should be considered, the former to accommodate hierarchical networks and the latter for flat network structures.
Objective 7: Determine the size of the coordination clusters in which NetMIMO systems should be organized.
Although, in its purest conceptual form, NetMIMO calls for each user being served by all the sites in the system, because of the power decay experienced by wireless signals as they propagate, in practice it suffices to serve each user via the sites within a certain range. This naturally leads to clustered NetMIMO systems, whose appeal is reinforced by the delay differences that the signals emanating from distinct sites accumulate over distance and which complicate the coordinated transmission/reception processes. Additionally, the formation of clusters has the benefit of reducing the demand for backhaul on the backbone network, a point that connects Objectives 6 and 7.