Section: New Results
Underwater acoustic communications
Participants : A. Kibangou [ contact person ] , C. Siclet, L. Ros, G. Gomez.
Nowadays, a great interest is dedicated to study fleets of AUVs working together to reach a common objective such as a gradient search for source detection; that is the overall goal of the CONNECT project. For this purpose derivation of distributed, multi-vehicle co-ordination schemes and development of efficient underwater acoustic communication protocols are needed. Current underwater acoustic modems are based on very classical single-carrier modulation with a very low bit rate. For achieving high data rate and large system capacity, Orthogonal Frequency Division Multiplexing (OFDM) has been claimed to be an efficient communication technology. It allows designing low complexity receivers to deal with highly dispersive channels. This fact motivates the use of OFDM in underwater environments.
Underwater acoustic channels are wideband in nature due to the small ratio of carrier frequency to the signal bandwidth, which introduces frequency-dependent Doppler shifts [65] . They also exhibit several propagation paths.
In order to adequately recover the transmitted information, algorithms at the receiver must include estimation and compensation of the Doppler scaling factor, channel estimation, and information symbols estimation. Several approaches have been suggested in the literature for estimating the Doppler scaling factor. They are based on the use of preamble and postamble of a packet consisting of multiple OFDM blocks [65] or by exploiting correlation induced by the cyclic prefix [64] . Then, the received signal is resampled by using a sampling period related to the estimated Doppler scaling factor. It is also necessary to estimate and compensate for the residual carrier frequency offset (CFO) since the Doppler can vary between consecutive OFDM blocks inside a given packet.
In our studies, the received data are processed block-by-block. We make use of high resolution harmonic estimation methods to estimate both Doppler scaling factor and channel parameters (path gains and delays) [62] . The advantage of the proposed scheme is to avoid data resampling and residual CFO estimation and compensation. The estimated channel parameters and Doppler are then used to build a channel matrix that is used for estimating the informative symbols following a zero forcing scheme [63] .
This work can be easily extended to the multiple access channel technique called OFDMA (Orthogonal Frequency Division Multiplex Access), which can significantly reduce the latency induced by TDMA (Time Division Multiplex Access) based protocols currently used. We also investigate transmitter/receiver using tensor algebra tools. We have indeed recently developed new algorithms [30] and applications on nonlinear system identification [31] , [14] based on multilinear algebra.
