Joydeep Acharya

Full-dimensional MIMO for future cellular networks

Full-dimensional multiple input multiple output (FD-MIMO) is a practical implementation of massive MIMO in cellular communication systems. FD-MIMO relies on an active two-dimensional antenna grid at the base stations and harvests the benefits of both azimuth and elevation beamforming. This paper briefly introduces the concept of FD-MIMO, and surveys the recent research results and applications of active antenna systems (AAS) technology. The key concepts discussed range from antenna patterns and channel modeling to vertical sectorization and heterogeneous networks deployment. The paper focuses on the benefits of FD-MIMO in an urban environment with high rises and three-dimensional user deployment. It demonstrates the importance of FD-MIMO in enhancing the performance of cellular networks.

Enhancing ZF-DPC Performance with Receiver Processing

Multiuser multiple-input multiple-output (MU-MIMO) systems are known to achieve superior data rates compared to single user (SU-MIMO) systems in the downlink. The achievable rates of MU-MIMO systems are dependent on the precoder design at the transmitter. Dirty paper coding (DPC) is a well known non-linear algorithm to achieve MIMO broadcast channel (BC) capacity. Zero forcing DPC (ZF-DPC) is a sub-optimal way to achieve DPC by triagularizing the channel. In a recent work, it was shown that ZF-DPC with receiver beamforming can provide further improvements. In this paper, we propose an enhancement of ZF-DPC with receiver processing by utilizing the MAC/BC duality. The proposed algorithm is iterative in nature which provides a useful trade-off between complexity and desired performance. Simulation results show that the proposed ZF-DPC algorithm achieves considerable improvement over conventional ZF-DPC and is within 1% of MIMO BC capacity.

Performance comparison of ZF-DPC to block diagonalization for quantized feedback

Multiuser multiple-input multiple-output (MU-MIMO) systems can achieve higher data rates compared to single user (SU-MIMO) by exploiting multi-user diversity. The capacity bound for MU-MIMO systems is given by the MIMO broadcast channel (BC) capacity. Dirty paper coding (DPC) is a well known non-linear algorithm to achieve this bound. Zero forcing DPC (ZF-DPC) is a sub-optimal way to achieve DPC by triagularizing the channel. In our previous work, we had proposed an enhancement of ZF-DPC by introducing additional receiver which was shown to improve the achievable rates over linear schemes such as block diagonalization. In practise Block Diagonalization is implemented by the receiver feeding back quantized information about the channel and the transmitter figuring out the precoders subsequently. To establish a comparison of ZF-DPC to Block Diagonalization for such practical feedback, we first derive the framework for operating ZF-DPC with quantized feedback. Next we compare the performance of ZF-DPC over Block Diagonalization with quantized feedback and demonstrate that the former still performs better.

Optimizing vertical sectorization for high-rises

Vertical sectorization is a recent development of active antenna technologies, whereas cellular providers make use of sectors in the vertical domain in addition to the standard horizontal sectorization currently in place. In this paper, we develop a framework to analyze and optimize vertical sectorization for high-rise buildings in urban environments. We start with a generic urban network topology with high-rise buildings and use it to derive the distribution of the elevation angles of the user equipments (UEs) in a cell. Using this distribution, we derive the values of tilts that maximize the coverage of all the UEs for the cases of one and two vertical sectors. This work provides answers to how a cellular service provider should perform vertical sectorization for different base station heights, cell sizes and various urban environments.

Multiuser MIMO Transmission in Distributed Antenna Systems with Heterogeneous User Traffic

In this paper, we consider multiuser transmission in the downlink of a distributed antenna system with two types of users: i) the real-time users that require a minimum instantaneous rate and ii) the best-effort users that try to maximize their rates. We start with a two-user system consisting of one real-time user and one best-effort user, in which the optimal beamforming scheme is proposed to maximize the instantaneous rate of the best-effort user under the rate constraint of the real-time user and per-antenna-power constraint. Two suboptimal transmission schemes are then proposed to reduce the computational complexity and feedback overhead, respectively. Furthermore, we propose a round-robin based scheduler to extend the proposed schemes to a general multiuser system, such that the outage probability of the real-time users is minimized. System-level simulation results show that the proposed schemes have better performance than two reference schemes.

Vertical beamforming for three-dimensional user placement

Full-dimensional multiple input multiple output (FD-MIMO) is a practical implementation of massive MIMO in cellular communication systems. FD-MIMO relies on a planar array of active antenna systems (AAS) at the base stations to harvest the benefits of both azimuth and vertical beamforming. In this paper, we analyze the network performance of per-user vertical beamforming with a special focus on the three-dimensional placement of the base stations and the users. We use tools from stochastic geometry to analyze the coverage performance of cellular networks with a three-dimensional radiation pattern at the base stations, using AAS. We show that the directionality achieved via dynamic antenna tilting provides multi-fold improvement in network performance as compared to omni-directional digital beamforming. We draw insights on the performance of the network as a function of the height difference between the base stations and the users, the number of antennas at the base station, and the various characteristics of the radiation pattern.

A Practical GDFE Precoder for Multiuser MIMO Systems

Generalized decision feedback equalizer (GDFE) achieves MIMO broadcast channel (BC) capacity but suffers from huge computational complexity and feedback overhead making it unsuitable for practical systems. Much of the complexity is due to the computation of the covariance matrix corresponding to the least favorable noise. Additionally GDFE implementation requires the base stations (BS) to feedback filter coefficients to the users thus increasing overhead. In this paper, we address both the aspects - computational complexity and feedback overhead. We first propose an alternative framework for realizing the GDFE precoder, which avoids the explicit computation of the least favorable noise. We show that, while maintaining capacity optimality of the GDFE precoder, the proposed algorithm has significantly lower complexity. Next, we propose another algorithm which does not require the feedback of filter coefficients from BS to users with minimal loss in achievable capacity. Thus our algorithms are a step towards adopting GDFE precoders in future wireless systems.

A Near-Capacity GDFE-Like Precoder with Reduced Feedback Overhead for MIMO Broadcast Channel

Downlink multi user multiple input multiple output (MU-MIMO) systems are of increasing importance in current and upcoming wireless applications. The improvement in data rates offered by such systems depends on the design of precoding schemes for the broadcast channel (BC) which is their theoretical generalization. A precoding scheme based on generalized decision feedback equalizer (GDFE) is known to achieve MIMO BC capacity. However, GDFE precoder suffers from huge computational complexity and feedback overhead that renders it unsuitable for practical systems. While the computational complexity has been dealt elsewhere, the feedback overhead issue has not yet been resolved. Thus in this paper we propose an algorithm to significantly reduce the feedback requirements of GDFE and show that the performance loss is minimal. We do this by first presenting a variant of the conventional GDFE algorithm, which we prove has the same theoretical performance.