It is well known that current reactive network management would be unable to support the exponential increase in complexity and rapidity of change in future cellular networks. Keeping this in perspective, the goal of this paper is to investigate applicability of machine learning and predictive models to assess cell-level user quality of experience (QoE) in real-time. For this purpose, we leverage a 5 week LTE metrics data collected at cell level granularity for a national LTE network operator. Domain knowledge is applied to assess user QoE with network key performance indicators (KPIs), namely scheduled user throughput, inter-frequency handover success rate and intra-frequency handover success rate. Results indicate that applying boosted trees model on a subset of carefully selected non-collinear features allows high accuracy threshold-based estimation of user throughput and inter-frequency handover success rate. We also exploit the periodic nature of cell data characteristics and apply a recently developed time series prediction model known as PROPHET for future QoE estimation. By employing machine learning and data analytics on network data within an end-to-end framework, network operators can proactively identify low performance cell sites along with the influential factors that impact the cell performance. Based on the root cause analysis, appropriate corrective measures may then be taken for low performance cell sites.

The Mobile Internet connects billions of smart phones and laptops. With this global connectivity, the stage is set for the emergence of: (a)Tactile Internetto deliver haptic experiences to remote users, and (b) flexible and integratedBackhaul/Fronthaulnetworks to support demands of such applications.

5G wireless networks are expected to usher in massive Machine Type Communications (mMTC) pertaining to the Internet of Things (IoT). A large fraction of the devices participating in IoT are likely to be low-power devices that generate sporadic traffic. It is anticipated that autonomous grant-free communications would be needed to accommodate these devices and the traffic they generate. Preamble transmission is a key part of such grant-free communications, and system performance including the desired power savings for the devices depends critically on the preamble detection-false alarm characteristics. In this context, low-power devices in extended coverage scenarios present a stiff challenge to system designers. In this study, we show that with narrowband resource allocations, use of preamble acknowledgements and judiciously set power control targets one can achieve the desired performance even for low-power devices in challenging scenarios.

Throughput performance of WiFi networks, which employ the Carrier Sense Multiple Access (CSMA) technique, can be estimated fairly accurately using Bianchi's analytic solution provided all elements of the network are within the carrier sensing range of one another. However, this solution cannot be applied to networks that are spread over a wide area. A paper by Liew, et al proposed an interesting approximate solution for such networks. While this solution was shown to work well in a variety of conditions, it suffers from a limitation in that it requires all transmitter-receiver pairs to have independent and identically distributed (i.i.d.) transmission times. In this paper, we generalize some of the key results of Liew, et al by removing the restriction on transmission times, which leads to a solution technique that can be used to analyze CSMA networks where different transmitter-receiver pairs operate at different bit rates. We also provide a few examples to illustrate how well this technique works in different settings and compare it with solutions based on other methods.

In this paper, we present a practical method to construct dirty paper coding (DPC) schemes using sum codes. Unlike the commonly used approach to DPC where the coding scheme involves concatenation of a channel code and a quantization code, the proposed method embodies a unified approach that emulates the binning method used in the proof of the DPC result. Auxiliary bits are used to create the desired number of code vectors in each bin. Sum codes are obtained when information sequences augmented with auxiliary bits are encoded using linear block codes. Sum-code-based DPC schemes can be implemented using any linear block code, and entail a relatively small increase in decoder complexity when compared to standard communication schemes. They can also lead to significant reduction in transmit power in comparison to standard schemes.

Base station cooperation has been shown to significantly mitigate the interference perceived in cellular networks and improve the quality of user experience. On the uplink, joint processing (JP) across antennas within a cluster of base stations yields significant gain, but requires significant backhaul to transport complex I & Q signals. The Network Interference Cancellation Engine (NICE) can yield throughput gains similar to that of JP, but with one to two orders of magnitude less backhaul. While NICE demonstrates significant potential for cellular networks, its performance has been evaluated under somewhat idealistic conditions. For instance, full buffer traffic has generally been assumed, thereby leading to a fully loaded system with strong interference. Bursty traffic may result in a partially loaded system with less interference and impact the gains of interference-mitigating techniques such as NICE. In addition, an ideal backhaul with negligible delay has generally been assumed. Delays in information exchanges between base stations may decrease the efficacy of base station cooperation. In this paper, the performance of NICE is evaluated under non full buffer traffic and non-ideal backhaul.

In this paper, we present a practical Dirty Paper Coding (DPC) scheme using sum codes based on LDPC codes. While a typical sum-code-based DPC scheme uses a constrained decoder as part of the encoding operation, such an approach fails in the case of LDPC codes since the constrained decoder based on a standard LDPC decoder tends to get stuck at highly suboptimal points in typical scenarios. We get around this difficulty via a simple code construction method based on bit-mapping or permutation. Examples included in this paper show that compared to standard communication schemes the proposed DPC scheme yields substantial savings in terms of the transmit energy required to attain desired error performance.

While the information theoretic result commonly referred to as “dirty paper coding” [1] has been known for long, it has not led to practical schemes that can be implemented easily in real-life communication systems. In this paper, we present the concept of sum codes and show how they provide a convenient platform for dirty paper coding. Comb-combined sum codes, which provide a convenient method of sum-code construction, are particularly useful in this respect. One can build comb-combined sum codes based on any linear block code to achieve reduction in the transmit energy required to attain a desired performance. Using an example based on tail-biting convolutional codes, we show how practical dirty paper coding may be implemented using any convenient linear block code.

Although the information theoretic result referred to as “dirty paper coding” [1] has been known for quite a while, it has not led to schemes that look attractive enough for real-life communication systems. In a companion paper [2], we presented the concept of sum codes and showed how they provide a convenient platform for interference suppression via dirty paper coding. In this paper, we present a dirty paper coding scheme based on trellis-coded modulation, which could be looked upon as an alternative to the sum-codes-based approach. Since trellis-coded modulation is designed for multi-level signal constellations, it is expected to work well as a platform for the proposed dirty paper coding scheme in which multi-level signal constellations play a critical role.

It is widely believed that cooperation among base stations can be employed to enhance the interference mitigation capability of base station receivers in cellular networks. Among the base station cooperation schemes that have been proposed to address the uplink interference problem, two promising ones are 1) Joint Multi-Cell MMSE Processing also referred to as Network MIMO; and 2) Multi-Cell Successive Interference Cancellation (MC-SIC). While both of these methods can substantially improve the performance of cellular networks, they have complementary strengths and limitations when implemented in a practical setting. In this paper, we propose a hybrid scheme called Generalized Network Interference Cancellation Engine, which combines the strengths of these two approaches while working within the constraints imposed by practical implementations. This view is borne out by the performance results presented in this paper.

The demand for high-rate wireless services grows unabated as new mobile devices and data-hungry applications continue to emerge. However, despite many advanced physical and link layer techniques that are employed in today's cellular networks, a large number of mobile users, especially at the cell edges, receive inadequate data rates due to high interference and path loss. Several techniques that exploit inter-base station cooperation have been proposed to address the inter-cell interference problem. While the initial studies show promising gains in idealized environments, it is important to analyze these techniques in a more realistic setting. This study provides detailed analyses of the benefits provided by uplink base station cooperation with several practical constraints and issues that would be encountered in a realistic cellular network deployment. Our results indicate that the cooperative signal processing techniques are robust to channel estimation errors given the pilot structure available in contemporary wireless systems. Furthermore, uplink cooperation can substantially improve the interference management capability of the cellular network, providing significant rate improvement over current non-cooperative networks.

Heterogeneous networks comprising overlapping layers of macro cells and small cells are expected to be widely deployed to cope with the rapidly increasing demand for wireless bandwidth. While judicious resource sharing and interference management mechanisms allow these networks to provide adequate performance for static or pedestrian users, they often fail to meet the needs of highly mobile users. In this paper, we propose a cooperative data transmission scheme called "Seamless Macro-Cell-Anchored Radio Transmission" that enables even highly mobile users to experience performance levels similar to those enjoyed by static users. The performance improvements effected by this scheme occur mainly at low percentiles; in that sense it can be said to deliver sustained end-user experience even to highly mobile users. Further improvement in edge performance is possible with the use of rate thresholds in conjunction with the proposed scheme.

Recently, there has been considerable interest in the potential of inter-base station cooperation techniques to manage excessive out-of-cell interference in evolving cellular networks. It has been observed that on the uplink, significant improvements in interference rejection can be attained by exchanging complex-valued received signal samples across a cluster of cells and employing joint multi-antenna receiver processing. However, such processing incurs substantial increases in backhaul overhead, exceeding the actual information delivered by at least an order of magnitude. In this paper, we propose and describe a signal processing approach based on a novel Network Interference Cancellation Engine (NICE), which opportunistically cancels out-of-cell interference by exchanging decoded dominant interferer data among cooperating cells and reconstructing interfering signals. Simulation results with NICE processing in cellular networks show that substantial improvements (5-7 dB) in signal-to-interference-plus-noise ratio may be achieved over current non-cooperative multi-antenna receiver processing techniques while allowing backhaul overhead to be reduced by factors on the order of 40 to 50 relative to joint processing techniques employing the exchange of complex-valued received signal samples.

Single frequency networks (SFN) in combination with a multi-carrier transmission scheme such as OFDM are often proposed as a means to provide continuous coverage of broadcast applications over a large area. In this paper, we propose a distributed beam-forming algorithm that makes use of channel state information fed back by receivers to improve the performance of the weaker connections. The distributed nature of the proposed algorithm makes it easier to implement when compared to its centralized counterparts. Simulation studies show that the proposed algorithm results in a significant improvement in the performance of SFNs supporting broadcast or multicast applications.

In this paper, the feasibility and capacity benefits of deploying low-cost relay nodes with in-band wireless backhaul capability in cellular networks are studied. Our proposed scheme is similar to a cell-splitting architecture where the base station density is increased, thus reducing the effective cell size. Cell-splitting in an interference-limited deployment provides capacity gains that are proportional to the number of additional base stations deployed, but at significant cost driven primarily by wired backhaul requirements. The use of in-band wireless backhaul may significantly reduce the cost, but sharing bandwidth with the backhaul results in reduced access rates relative to that of cell-splitting with wired backhaul. Further cost savings may be achieved with the use of low power, low complexity relays in lieu of the additional base stations. In this paper, we show that a cellular architecture with base stations supplemented by simple decode-and-forward relays can provide significant capacity improvements despite the overhead associated with in-band backhaul.

Single frequency networks (SFN) in combination with a multi-carrier transmission scheme such as OFDM are often proposed as a means to provide continuous coverage of broadcast applications over a large area. While this combination does succeed in providing continuous coverage, it has certain limitations. Specifically, at points where signals from multiple base-stations are received, the overall signal-to-noise ratio can be poor because of relative phase differences. In this paper, we show how, through the use of orthogonal space-time codes, this limitation of SFNs can be overcome to improve their overall performance. This scheme could be viewed as an ldquoopen-looprdquo implementation of multi-cell cooperative transmission.

In this paper, we investigate joint resource allocation and routing for a wireless mesh network consisting of fixed access points inter-connected through multi-channel wireless links. Some of the mesh routers are assumed to function as gateways with high bandwidth connections to a wired network and the mesh network is assumed to provide multi-hop capability where the traffic entering the mesh via gateways may be carried over multiple wireless hops towards the destination mesh routers. The main objective of this work is to optimize resource allocation and routing in such networks; in this regard, we study a cross- layer optimization problem that includes power control, channel allocation, link scheduling and routing. We provide results highlighting the capacity benefits of our proposed approach; these results indicate that significant throughput improvements can be achieved with cross-layer system design.

In this paper, we propose a cross layer optimization framework for multi-hop routing and resource allocation design in an orthogonal frequency division multiple access (OFDMA) based wireless mesh network. The network under consideration is assumed to consist of fixed mesh routers (or base station routers) inter-connected using OFDMA wireless links with some of the mesh routers functioning as gateways to a wired network. The objective of our cross-layer formulation is to allow joint determination of power control, frequency-selective OFDMA scheduling and multi-hop routing in order to maximize the minimum throughput that can be supported to all mesh routers. Results of our investigations under typical cellular deployment, propagation and channel model assumptions show that this approach achieves significant mesh throughput improvements primarily due to the following: (a) frequency selective scheduling with OFDMA which provides improved tone diversity thus allowing more efficient bandwidth utilization relative to single carrier methods; and (b) multi-hop routing which provides improved path diversity relative to single hop transmissions.

This paper investigates the performance benefits of different resource allocation strategies that take advantage of permutation zones specified in IEEE 802.16e-based OFDMA systems. In particular, we propose and investigate the benefits of a sub-zone scheduling approach where each permutation zone is partitioned into sub-zones that are used for scheduling subsets of eligible users according to their prevailing link quality. In the scenarios considered, the proposed sub-zone scheduling approach allows improvements in sector throughputs as high as 80% relative to a baseline IEEE 802.16e-based OFDMA system while maintaining the system outage at acceptable levels.

In this paper, different neighbor discovery methods are considered for adhoc networks without fixed spectrum allocations where discovery of neighboring nodes may need to occur over a large frequency range (e.g. tens of MHz to a few GHz). In such systems, co-existence policies or dynamic spectrum access protocols may also introduce temporal variability in the availability of frequencies for neighbor discovery thus making it challenging to discover neighbors in a short period of time and with low energy consumption. In this paper, we analyze the performance of a baseline scheme (representative of current techniques) where nodes perform discovery by transmitting and receiving beacons randomly across a large frequency range. We also propose and analyze a family of fast, energy-efficient neighbor discovery techniques that use pseudo-random frequency hopping sequences for beacon transmission and reception. For the cases considered, these techniques provide substantial improvements in discovery time and energy consumption relative to the baseline

An adaptive sleeping and awakening protocol (ASAP) is proposed for nodes in a synchronous adhoc sensor network. In order to increase energy efficiency, nodes within the network enter into a sleep mode and awaken at pre-determined time slot(s) to listen for transmissions from its immediate neighbors. Implicit knowledge of awakening slots for neighboring nodes is used to schedule transmissions within the neighborhood. Finally, nodes adapt their sleeping cycles based on neighbor topology and remaining battery life in order to maximize the network lifetime while satisfying the latency requirements of the underlying sensor application. Simulation results show that with delay constrained routing, ASAP can achieve twice the energy efficiency (or battery life) of a synchronous awakening approach.

A new class of channel assignment methods are investigated in a power controlled, time slotted wireless communications system. In practical systems, there are constraints on the transmit power. Channel assignment algorithms that do not consider these constraints achieve smaller power control gains and result in increased co-channel interference and reduced system capacity. It is shown that assignment algorithms that employ users' received signal strength provide significant gains in the presence of these constraints.

The performance of turbo codes is investigated in a frequency-hopped spread-spectrum (FH-SS) system with full-band thermal noise and Rayleigh fading. For cases where the data rate exceeds the hopping rate (i.e., there exists multiple bits per hop) and fading side information (SI) does not exist, our approach is to iteratively estimate the fading levels. Simulation is performed for coherent and noncoherent reception, variable number of bits per hop, and cases where fading SI is available or unavailable to the decoder. It is shown that iterative channel estimation performed in conjunction with iterative decoding can improve the overall decoding performance. Finally, the performance of an FH-SS system using standard fading assumptions is compared to the performance of a measured fading channel. Due to flaws in the assumptions, the measured fading channel yields several decibels of performance loss relative to the ideal fading channel.

Turbo codes are investigated in a slow frequency-hopped spread spectrum (FH-SS) system with partial band jamming. In addition, full-band thermal noise is present. The channel model is that of a partial-band jammer in which a fraction of the frequency band is jammed and the remaining fraction is unjammed. This paper focuses on the implementation and performance of a modified turbo decoder for this model. We refer to the knowledge that each transmitted bit is jammed as channel state information. We consider cases of known or unknown channel state and variable number of bits per hop. Our approach is to modify the calculation of branch transition probabilities inherent in the original turbo decoder. For the cases with no side information and multiple bits per hop, we iteratively calculate channel state estimates. Analytical bounds are derived and simulation is performed for noncoherent demodulation. The performance of turbo codes is compared with a Reed-Solomon and a concatenated code comprised of a convolutional inner code and Reed-Solomon outer code.