Model-Data Fused Estimation And Equalization For Time-Varying Shallow Water Acoustic Channels

Acoustic communications is a major tool for wireless data transmission in the un-derwater environment. Underwater acoustic (UWA) propagation is mainly restricted by three factors:attenuation increased with frequency, time-varying multipath, and relatively low sound speed. Due to these propagation properties, the UWA channel, especially the shallow-water acoustic channel is characterized as bandwidth-limited and severely delay-Doppler-spreading. Thus, the UWA communications faces numerous technical challenges different from radio communications in air.Aiming at reliable acoustic communications in a time-varying shallow-water envi-ronment, this thesis conducts a research incorporating channel estimation and equaliza-tion. Exploiting the sparse channels, compressed sensing methods are adopted to improve the channel estimation performance and reduce the complexity. Especially, time-reversal (TR) processing and frequency-reversal processing, which is a dual to TR in frequency domain, are introduced to facilitate the channel equalization. Besides, we extend the study of channel estimation and equalization from single-input single-output (SISO) systems to multiple-input multiple-output (MIMO) systems. Based on the theoretical research, we have also designed the communication experiment system and conducted several field ex-periments in the lake and in the sea. The results of experimental data processing have verified the effectiveness and performance of the schemes proposed in this thesis.The UWA channel is the medium of sound propagation. The study on the channel characters is the basis for the analysis and design of communication systems. The UWA channel, especially the shallow-water acoustic channel is one of the most complex chan-nels. The characters of UWA channels are determined by the sound propagation properties. The multipath propagation and Doppler effect make the shallow-water acoustic channel a typical delay-Doppler doubly spread channel. This thesis develops a channel model that can describe the doubly spread acoustic channel. Meanwhile, exploiting the sparsity of the channel, the UWA channel can be well approximated by several dominant discrete paths. Then, we adopt two polynomials to fit the amplitude variation and delay variation of each path, respectively. Under such channel parameterizations, a mathematical model for the discrete-time channel input-output relationship is derived. The derived channel model that parameterizes the time variation of each path takes into account of some issues neglected by other models such as the path amplitude variation and the deviation of the sampling points on the pulse shaper due to delay variation. Thus, the channel model proposed in this thesis can approximate the time variations of practical channels better.The parameterization channel model mentioned above uses a few parameters to rep-resent the channel input-output relationship, transforming the channel estimation prob-lem into the estimation of the low-dimensional parameters. Based on the parameterized channel model, we adopt compressed sensing methods to reduce the complexity of chan-nel estimation. Noting that the equivalent channel after resampling and carrier-frequency-offset (CFO) compensation changes much more slowly, a two-stage channel estimation approach which estimates the delay and Doppler scale sequentially is proposed. The two-stage approach degrades the search area from two-dimensional delay-Doppler scale plane to unidimensional delay/Doppler grid. Thus, the computational complexity and memory requirement are lower due to reduced dictionary size. Moreover, the two-stage approach can avoid the coupling between parameters, resulting in more accurate estimates. Both nu-merical simulations and experimental data processing have validated the performance of the two-stage approach.The severe inter-symbol interference (ISI) caused by the delay spread in shallow-water environment is a substantial limitation for UWA communications. To solve the ISI problem, we investigate the channel equalization in UWA communications. Time reversal (TR), which exploits the environment itself to compress the extended signal, is a signal processing method based on the physical properties of the environment. Though TR pro-cessing cannot completely eliminate the ISI, it can greatly simplify the design of equalizer to incorporate TR processing with channel equalization. This thesis applies estimates of the time-varying channel based on the parameterization channel model in TR processing, extending the conventional TR to time-varying channels. Besides, this thesis proposes the concept of frequency reversal dual to TR. Frequency reversal can compress the frequen-cy spread caused by Doppler to eliminate the inter-carrier interference (ICI) in orthogonal frequency division multiplexing (OFDM) communications.This thesis further considers the problem of channel estimation and equalization in MIMO UWA communication systems. The parameterization of path time-variation is ex-tended to MIMO UWA channels, based on which a discrete-time mathematical representa-tion of the input-output relationship for MIMO channel is derived. The two-stage approach is also extended to the estimation of MIMO channel. In the study of channel equaliza-tion for MIMO channels, we combine TR with MIMO. Integrated with MIMO, TR can implement multiuser communications due to its spatial focusing character. The classic de-cision feedback equalizer (DFE) in SISO is extended to MIMO DFE, which can eliminate both the ISI and the cochannel interference (CCI) from other users simultaneously. Be-sides, frequency domain equalization based on frequency reversal is extended to MIMO communications. Similar to TR, the frequency reversal can achieve spatial focusing while eliminating the CCI, thus providing a potentially effective frequency domain equalization approach in MIMO OFDM.