Soft-Output Joint Channel Estimation and Data Detection using Deep Unfolding

Abstract

We propose a novel soft-output joint channel estimation and data detection (JED) algorithm for multiuser (MU) multiple-input multiple-output (MIMO) wireless communication systems. Our algorithm approximately solves a maximum aposteriori JED optimization problem using deep unfolding and generates soft-output information for the transmitted bits in every iteration. The parameters of the unfolded algorithm are computed by a hyper-network that is trained with a binary cross entropy (BCE) loss. We evaluate the performance of our algorithm in a coded MU-MIMO system with 8 basestation antennas and 4 user equipments and compare it to state-ofthe-art algorithms separate channel estimation from soft-output data detection. Our results demonstrate that our JED algorithm outperforms such data detectors with as few as 10 iterations.

Simple and efficient visualization of aromaticity: bond currents calculated from NICS values

“…We see many open research problems. First and foremost is the design of techniques that further reduce the complexity of our JED algorithm-a promising direction is our recent work in [76] that solves a related JED problem in only 10 iterations with the aid of a neural network. Furthermore, methods as in [20]- [22], [77], [78] that decentralize data detection algorithms, so that complexity can be off-loaded to the APs and interconnect data rates can be reduced, will be key for a practical deployment of JED-here, the formation of overlapping UE clusters might be beneficial.…”

Mitigating Smart Jammers in Multi-User MIMO

“…We support the soundness of the proposed optimization problem by proving that its global minimum is unique and recovers the transmitted data symbols, given that certain sensible conditions are satisfied. By building on techniques for joint channel estimation and data detection [27]- [33], we then develop two efficient iterative algorithms for approximately solving the optimization problem. The first algorithm is called MAED (short for MitigAtion, Estimation, and Detection) and solves the problem approximately using forward-backward splitting (FBS) [34].…”

“…The second algorithm is called SO-MAED (short for Soft-Output MAED) and extends MAED with a more informative prior on the data symbols to produce soft symbol estimates. SO-MAED also relies on deep unfolding to optimize its parameters [33], [35]- [38]. We use simulations with different propagation models to demonstrate that MAED and SO-MAED effectively mitigate a wide variety of naïve and smart jamming attacks without requiring any knowledge about the attack type.…”

“…To improve detection performance, especially in the cases where MAED performs suboptimally, we now provide a second algorithm for approximately solving the problem in (11). This second algorithm builds on MAED but replaces the proximal operator in (22), which enforces MAED's box prior, by an approximate posterior mean estimator (PME) based on the discrete symbol prior as in [33]. Since the PME also enables meaningful soft-output estimates of the bits that…”

“…Based on this idealized model, we utilize a three-step procedure as in [33] for computing symbol estimates. First, we use (30) to compute log-likelihood ratios (LLRs) for every transmitted bit.…”