| dc.description.abstract | Signaling overhead reduction has always been a challenging task in the design of wireless networks. New applications and services in future networks, such as autonomous vehicles, and deployment of higher frequency bands, e.g., millimeter-wave and terahertz bands, face the common issue of wireless channels with large number of paths that have time-varying coefficients. The rich multipath components of the channels create delay spread. It causes inter-symbol interference in the signal, which has to be managed by guard periods in time, e.g., cyclic prefix (CP). Additionally, a high-mobility wireless environment between the transmit and receive antennas causes Doppler shifts in the channel, resulting in time-varying channel coefficients. In such scenarios, frequent channel estimation updates are required, demanding frequent transmission of pilot sequences. Consequently, the pilot signal and CP lead to substantial signaling overhead that negatively impacts spectral efficiency and leads to additional latency. Therefore, the focus of this thesis is on the overhead reduction. In particular, this thesis proposes novel approaches, for contemporary and future wireless networks, capable of dealing with the delay and Doppler spreads of the channel with significantly reduced overhead.
The first contribution of this thesis reduce the overhead of one of the main technologies of 5G systems, the orthogonal frequency-division multiplexing (OFDM)-based massive multiple-input multiple-output (MIMO).
This thesis brings substantial training overhead reduction through a novel joint channel estimation and equalization technique that requires only a single pilot subcarrier. This is possible by exploiting the channel correlation property induced by spatial diversity, and the fundamental concept of coherence bandwidth. Given the design parameters of 5G systems, the coherence bandwidth extends across multiple subcarrier bands. Hence, a band of subcarriers can be equalized with the channel estimate at a single subcarrier. Subsequently, the detected data symbols are considered as virtual pilots, and the channel frequency responses (CFRs) are updated at each subcarrier. Thereafter, the remaining channel estimation and equalization can be performed in a sliding manner. Furthermore, multiple channel estimates within the coherence bandwidth create additional frequency diversity, therefore multiple channel estimates are used to equalize the data at each subcarrier. This makes the proposed technique surpass contemporary systems that use multiple pilot subcarriers in terms of bit error rate (BER) and signal-to-interference-plus-noise ratio (SINR) performance.
As the new generation of applications and services emerge, the wireless environment becomes more time-varying, and hence, a larger overhead is required. In OFDM, the Doppler shifts of the channel break the orthogonality among the subcarriers, leading to severe inter-carrier interference. As a paradigm-shifting technology, orthogonal time frequency space (OTFS), which places the data symbols in the delay Doppler (DD) domain, has recently emerged. In the DD domain, the equivalent channel is sparse and time-invariant. As the second contribution, this thesis uses time-reversal maximum ratio combining (TR-MRC) to completely remove the CP of OTFS-based massive MIMO systems, substantially reducing the overhead. Noticeably, apart from the SE gains achieved by removing the CP, the proposed technique also reduces the duration of the OTFS block, which mitigates the effects of channel aging. Additionally, the asymptotic analysis in this thesis reveals that TR-MRC averages the large variations of the channel into the channel correlation. This insight leads to breaking the limitations of OTFS with the proposed residual Doppler correction (RDC) windowing. The specific limitation addressed is the assumption of approximately constant channel gains for the delay blocks of samples within each OTFS block. Simulation results show that using the RDC windowing to correct the channel variations within each OTFS block leads to a SINR performance close to static scenarios, even for substantially larger Doppler spreads.
Another issue of OTFS systems is the large spectral efficiency loss caused by the overhead required for channel estimation. In the current literature, OTFS deploys mostly embedded pilot (EP) structures, which use guard intervals that are twice the length of the channel delay spread to avoid interference between the pilot and data symbols. Hence, to reduce the pilot overhead, this thesis proposes a novel split pilot structure with two impulse pilots that share the same reduced guard interval. This way, the proposed split pilot can reduce the channel estimation overhead by almost half compared to the EP scheme. The challenge to reducing the delay guard rests on the interference between the pilot and the data symbols. With two impulse pilots, each pilot can be used to remove the other pilot's interference over data. This is while the data interference over the pilot can be initially absorbed into noise due to the large pilot power. To improve the accuracy of this method, this thesis also proposes to remove data interference over the channel estimates iteratively using the initial data estimates. This significantly improves the BER performance of split pilots with as few as two interference cancellation stages. | en |
