Transmission strategies under imperfect instantaneous CSIT

Abstract

We propose a simple solution based on PerronFrobenius theorem and uplink-downlink duality to the longstanding problem of transmission strategies: maximize the minimum SINR or minimize the transmit power. In contrary of using perfect and instantaneous CSIT (requires higher overhead and backhauling), or using only average CSIT, e.g. channel covariance which suffers from quality fluctuations; we account for imperfect instantaneous CSIT. The solution emerges from the observation that the crux of the robustness problem is the construction of SINR estimates itself. We provide a MLE-based estimate which being sub-optimal and failing to incorporate the channel estimation error, we reinforce it by biasing the estimate, which not only achieves smaller error but is strictly conservative as well. We also consider both transmission strategy problems, and show that the biased estimates are not only robust compared to traditional unbiased estimates, but also substantially outperform the probabilistic method with Bernstein-type relaxation proposed in [1].

Full-Duplex Transmission Optimization for Bi-directional MIMO links with QoS Guarantees

“…The optimization problem described by equation (7) is wellknown to be non convex due to the SINR constraints [23], although approximate solutions can be obtained for it with basis on convex optimization algorithms, such as interior point methods if the constraint can be convexified [24]. In addition to the losses due to convex relaxation, such solutions tend also to be computationally demanding.…”

Multi-Cell MIMO User Rate Balancing with Partial CSIT

“…Then, an iterative algorithm is proposed to solve the expected MSE balancing problem by switching between the broadcast and the multiple access channels. Also, SINR balancing problem with imperfect CSIT is studied in [16] for multi-cell multi-user MISO system. Therein, the authors introduce an alternative biased SINR estimate to incorporate the knowledge of the channel estimation error, outperforming the unbiased maximum-likelihood estimate.…”