| dc.description.abstract | Conventionally, information bits are conveyed in a communication system by transmitting symbols from a complex modulation alphabet such as QAM or PSK. The wireless fading channel is viewed as a signal distorting medium which changes the amplitude and phase of the transmitted complex symbols in a random manner. A recently proposed modulation technique called `media-based modulation' (MBM) takes a different approach in which the channel fades from the transmitter to the receiver are used as a channel alphabet to convey additional information bits along with the bits conveyed by the conventional complex modulation symbols. MBM uses digitally controlled parasitic elements called as radio frequency (RF) mirrors placed near the transmit antenna to create the channel alphabet. Each RF mirror can be in ON or OFF state. An RF mirror reflects the incident RF signal when it is in the ON state and allows it to pass through when it is in the OFF state. The ON/OFF status of the RF mirrors is called as `mirror activation pattern' (MAP). MBM has been shown to achieve high rates and significant performance gains compared to conventional modulation schemes. In this thesis, we investigate several key issues in MBM including design of efficient signal sets for MBM, the effect of IQ imbalance on MBM, deep learning-based receiver architectures for MBM, design and analysis of novel transmission techniques using MBM, and use of MBM in uplink massive MIMO. The contributions made in this thesis can be summarized as follows.
Efficient signal set design for MBM:
Here, we address the problem of efficient signal set/constellation design for MBM. Multidimensional constellations have been extensively studied in the literature in the context of multidimensional coded modulation and space-time coded MIMO systems, where such constellations are formally called lattice codes. The constellation design for MBM is fundamentally different from those for
multidimensional coded modulation and conventional MIMO systems, mainly because of the inherent sparse structure of the MBM signal vectors. Specifically, we need structured sparse lattice codes with good distance properties. We show that using an (N,K) non-binary block code in conjunction with lattice-based multilevel squaring construction, it is possible to systematically construct a signal set for MBM with a certain guaranteed minimum distance. The MBM signal set obtained using the proposed construction is shown to achieve significantly improved bit error performance compared to the conventional MBM signal set. In particular, the proposed signal set is found to achieve higher diversity slopes in the low-to-moderate SNR regime.
Effect of IQ imbalance on MBM:
In this part of the thesis, we focus on a practical issue that arises in MBM systems which use direct conversion radio frequency (RF) front end architecture. Specifically, we study the performance of MBM in the presence of transmitter and receiver side IQ imbalance (IQI) and propose efficient compensation techniques to alleviate it. We derive the MBM system model in the presence of transmit and receive IQI and show by numerical simulations that MBM is more resilient to IQI compared to conventional modulation. We also propose a scheme that jointly estimates and compensates the channel and IQI parameters during the channel estimation phase using widely linear least squares (WLLS) technique. The proposed scheme is shown to alleviate the degradation in the bit error performance caused by transmit and receive IQI.
DNN architecture for MBM receivers:
Practical communication systems can have non-idealities which may not follow known models or which may lead to models for which obtaining optimal analytical solutions may not be feasible. Recently, deep learning (DL) based techniques have been successfully employed in addressing such problems in communications. In this direction, we propose a deep neural network (DNN) architecture for MBM receivers, which uses small sub-DNNs to detect the mirror activation pattern and the complex constellation symbol transmitted by the antenna. The proposed DNN-based detector is shown to outperform the conventional maximum likelihood detector under the following conditions of practical interest: i) correlated noise across receive antennas (resulting from mutual coupling, matching networks), ii) noise distribution deviating from the standard AWGN model, and iii) IQ imbalance at the transmitter and receiver. The proposed DNN-based detector learns the deviations from the standard model and thus alleviates the degrading effects of the non-idealities.
Multidimensional index modulation using MBM:
Next, we propose several transmission schemes using MBM, in which MBM is combined with other index modulation (IM) schemes to achieve indexing across multiple dimensions. The proposed schemes are collectively called as multidimensional index modulation schemes. The intuition behind these schemes is that to achieve a certain transmission rate, the proposed schemes convey a major part of the information bits through the indices of the transmission entities (time-slots, RF mirrors, and antennas) and thus allow using lower-order conventional modulation (QAM or PSK) alphabet, resulting in improved performance. We carry out the diversity and capacity analyses for one of the proposed schemes, viz., time-indexed MBM (TI-MBM). Our diversity analysis shows that TI-MBM can potentially extract the full channel diversity when used in multipath channels. Also, we derive the capacity achieving input distribution for the TI-MBM scheme and obtain the the probabilities with which time-slots and RF mirrors in TI-MBM can be activated such that the achievable transmission rate is maximized. Further, recognizing the sparsity which arises naturally in multidimensional IM systems using MBM, we propose low-complexity sparsity exploiting detection algorithms for the proposed multidimensional IM schemes. The proposed algorithms are shown to achieve good performance at low computational complexities.
MBM in uplink massive MIMO:
Finally, we consider uplink communication in a massive MIMO system consisting of tens of users and a base station (BS) with tens to hundreds of receive antennas. Each user employs MBM with one transmit antenna and multiple RF mirrors placed near it. Our results show that MBM, when used in the uplink massive MIMO setting, achieves superior bit error performance compared to other single RF chain based transmission schemes such as conventional modulation (using a single transmit antenna) and spatial modulation. We then propose a detection algorithm based on compressive sensing (CS) that exploits the inherent inclusion-exclusion sparsity of the MU-MBM transmit vectors. Our results show that the MU-MBM system, with the proposed CS based detection algorithm can lead to a significant reduction in the required number of receive antennas at the BS compared to uplink systems which use conventional modulation schemes. | en_US |
