A. Channel Representation, Estimation and Precoding for Correlated MIMO Systems

l   We propose efficient multi-dimensional nonlinear regression based channel representations which unify existing MIMO channel models and are general enough to describe various channel conditions and operation scenarios. More importantly, they offer significant reduction in the number of characterization parameters in many practical applications, especially for medially to highly correlated MIMO environments which are likely resulted from the use of large scale arrays. For a large scale MIMO system, the rank reduction effect is particular impressive. Our models thus enable significant lessening of the computation complexities associated with channel estimation, codebook design, codeword selection and feedback while maintaining good CSI and precoding qualities.

l   The reduced-rank CSI representation is very useful for many post-channel-estimation operations that require processing the instantaneous channel matrices. Depending on the specified modelling order, the proposed channel estimators offer tradeoff between identification accuracy and computational complexity. Moreover, the dimension-reduction induced noise rejection effect enables the proposed model-based estimators to achieve superior mean squared error (MSE) performance over certain SNR region when compared with that of the conventional approaches.

l   We developed a model-based closed-loop MIMO system. By incorporating our channel representation into the precoder design, the resulting precoded system provides significant reductions on the feedback bandwidth and the computational complexity needed for constructing the precoder and equalizer matrices. A very efficient codeword selection algorithm is also made possible. Numerical results show that compared with the conventional approaches that need full knowledge of instantaneous CSI, our proposal suffers only negligible performance degradation at very high SNR region.


B. Resource Allocation and Management for OFDMA and MIMO-OFDMA Networks

Algorithms for finding suboptimal and optimal solutions to total power minimization or capacity maximization resource allocation problems in OFDMA-based networks have been studied by many authors. But the complexities of finding the optimal solution and some suboptimal are prohibitively high and only few numerical examples for low dimension cases can be found. On the other hand, low-complexity suboptimal solutions often give unsatisfactory performance.

l   We propose optimal and suboptimal resource allocation solutions for OFDMA and MIMO-OFDMA wireless networks. Various design criteria and system constraints are considered. The corresponding complexities for all suboptimal solutions are relatively low while that for the optimal algorithm is only moderate high. We first investigate the problem of transmit power minimization in an OFDMA downlink network subject to user rates and BER requirements constraints. We provide near-optimal and optimal solutions based on the dynamic programming and branch & bound methodologies. The second scenario we consider is a weighted sum rate maximization problem which is solved via dynamic programming and dual decomposition.

l   We studied the product rate maximization scenario and presented a suboptimal solution. Finally, we consider a total power minimization problem for a multi-hop MIMO-OFDMA wireless network that employs fractional frequency reuse and present a low-complexity solution.

l   The main concept in our proposed algorithms can be easily applied to obtain a near-optimal solution for many similar multi-constraints optimization problems with low complexity.

C. Advanced Error-Control Schemes and Sequence Design

Error-control coding has become an indispensable part of a modern communication/storage system. With the proliferation of iterative decoding, the underlying “turbo principal” (belief propagation) has since crossed the borderline of coding theory and made far-reaching impacts on various scientific disciplines like communications, computer science, statistics, physics and bioinformatics, to name just a few. Similarly, various optimization and computational techniques used in other fields have been brought into play in the codec design. On the other hand, high rate applications require that a hardware-software (algorithmic) co-design approach for new generation of high throughput codecs. Recently, we have

l Presented several novel stochastic decoding algorithms for short linear block codes with high-density parity-check (HDPC) matrices. Our approach can be regarded as a randomized sphere decoding with moving center that selects candidate codewords around a center vector according to a sphere-symmetric probability distribution. The center (median) vector of the distribution is updated according to Monte Carlo based approach called the Cross-Entropy (CE) method so that the parameters of the updated distribution guarantee that the new distribution is closest to the optimal distribution in the sense of the Kullback-Leibler distance (i.e., CE).

l   Develop a new class of bit-flipping decoding algorithms for low-density parity-check codes. By adjusting the reliability of each reliable checksum and ignoring unreliable ones, we are able to detect (and flip) the unreliable bits with higher probability and bring the performance closer to that of the sum-product algorithm. The measurement for checksum reliabilities is improved by taking those of variable nodes (the value of the flipping function) into account. The proposed algorithms provide a noticeable gain over the known BF algorithms with limited increase in computational complexity.

l   Proposed a new bit-by-bit puncturing pattern searching scheme for LDPC codes. We also take into account the detrimental effects on previous punctured and unpunctured bits brought about by the new selected punctured. Given the bit locations which have been punctured, a new one is chosen from the set of candidate bits by i) examining its recovery capability (which depends on the number and reliabilities of its connected check node message) and ii) assessing the impact a candidate bit may make. Numerical experimental results show that the proposed scheme outperforms existing puncturing methods. The superiority and robustness of our scheme are further verified by some observed statistics and are consistent with a Gaussian approximation based analytic prediction.

Communication and radar systems called for special sequences that possess desired auto-correlation and cross-correlation properties for applications such as synchronization, channel estimation, pulse-compressed detection and implementation impairments compensations, to name a few. We have

l Developed three classes of new systematic approaches--a transform domain method and two direct (time domain) synthesis methods—for generating zero-correlation zone (ZCZ) sequences. The former approach generates sequence families that achieve the upper bounds on the family size and the zero-correlation zone width for a given sequence period. In addition, each generated sequence is a perfect sequence that has an ideal auto-correlation (AC) function. The latter two unify many existing construction methods and all of them provide more flexibility in producing new families of sequences.

l   Proposed a single-symbol based initial ranging signal structure and the associated signal detection and timing synchronization methods for OFDMA systems. Our structure makes it feasible and flexible for a receiver to detect single and multiple ranging codes and estimate the corresponding timing offsets. Our approach offers improved performance and enhanced robustness against multi-user interference and multi-path fading. Numerical results verify the effectiveness of the proposed method and its advantages over existing alternatives.

D. Estimation, Detection and Smart Transmissions

Data detection or fusion based on outputs from multiple wireless links often requires channel side information (CSI) about the links' error rates (ERs) performance.

l   We consider the scenario that these links include direct source-destination (SD) links and two-hop links that require an intermediate Decode-and-Forward node to relay the source signal. Conventional destination based estimators suffer from slow convergence and are incapable of simultaneously blind-estimating all ERs including, in particular, those of the source-relay (SR) links. They may also require various degrees of CSI about the ERs of the SD and relay-destination (RD) links to remove the ambiguity arose from insufficient number of links in the network and from that due to the symmetric nature of a cascaded source-relay-destination (SRD) link's ER as a function of its component SR and RD links' ERs. We proposed novel Monte-Carlo based estimators that overcome all these shortcomings. The estimation process involves injecting noise into the samples received by the destination node to create virtual links and alter link output statistics. We show that the latter scheme exhibits a stochastic resonance effect, i.e., its mean squared estimation error (MSEE) performance is enhanced by injecting proper noise and there exists an optimal injected noise power level that achieves the maximum improvement. The stochastic resonance effects are analyzed and numerical examples are provided to display our estimators' MSEE behaviors and to show that the ER performance of the optimal detector using the proposed estimators is almost as good as that with perfect ER information. The binary case is further extended to the case when sensors generate M-ary signals to the fusion center.

Spatial modulation (SM) which carries part of the information in the antenna index is an interesting multi-antenna based transmission scheme. It induces no inter-channel interference and requires only a single transmit RF chain thus greatly simplify the transceiver complexity and system overhead.

l  We propose a spatial modulation SM system using co-located dual-polarized antennas arrays. The use of a co-located dual-polarized array is due to the fact that it is more space- and cost-effective than a conventional array. Two suboptimal detectors which need lower complexity than the ML detector is proposed. These detectors outperform the matched-filter (MF)-based method and give near-ML performance. As the errors associated with antenna index detection and regular demodulation are different, we analyze the BER performance and suggest an unequal error protection coding scheme to enhance the system performance.