This paper considers network capacity and user coverage improvement in Internet of Things (IoT)-oriented massive MIMO systems. In the literature, user grouping approaches have been used in massive MIMO to improve the network capacity, where users are generally divided into non-overlapping groups, and those users with less favorable channel conditions are dropped for capacity optimization. As a result, users may suffer from unpredicted interruptions and delays, even long time disconnection from the network. Moreover, non-overlapping user grouping also leads to unnecessary resource waste. As an effort to overcome these limitations, in this paper, we introduce the concept of overlapping user grouping by exploiting the favorable propagation property in massive MIMO. More specifically, we propose two new user grouping approaches. First, we present a greedy search-based user grouping method by allowing overlapping among the selected subgroups. Second, we introduce a new channel similarity measure, and develop a low complexity overlapping user grouping approach based on the spectral clustering algorithm in machine learning. Both the theoretical and numerical results demonstrate that: overlapping user grouping can achieve much higher network capacity, and can ensure that at any given time, each IoT device will be served in at least one subgroup.
This paper considers physical layer interference management and network capacity improvement in Internet of Things (IoT) oriented cognitive MIMO-OFDM systems, where the IoT devices access to the licensed spectrum by cognitive radio in spectrum underlay. Aiming to fully eliminate the interference to primary user as well as ensuring optimal QoS for all the cognitive IoT devices, in this paper, we propose a leakage-based precoding scheme with dimensionality reduction and SLNR criterion. More specifically, we first introduce a new channel estimation method using the Multi-Channel Ratio (MCR) algorithm that requires no feedback from the primary network. Second, with the estimated interference channel, we eliminate the interference to the primary user by a low complexity subspace projection with dimensionality reduction. Finally, by fully exploiting the generalized Rayleigh quotient, we propose the leakage-based precoding scheme that aims to optimize both the leakage power and noise power of the IoT devices. Theoretical and numerical results demonstrate that the proposed leakage-based precoding scheme can satisfy the low-cost and self-organized features of IoT networks, and achieves significant improvements on BER performance and network capacity while causing no interference to the primary user.
This paper considers interference management and capacity improvement for Internet of Things (IoT) oriented two-tier networks by exploiting cognition between network tiers with interference alignment (IA). More specifically, we target our efforts on the next generation two-tier networks, where a tier of femtocell serving multiple IoT devices shares the licensed spectrum with a tier of pre-existing macrocell via a cognitive radio. Aiming to manage the cross-tier interference caused by cognitive spectrum sharing as well as ensure an optimal capacity of the femtocell, two novel self-organizing cognitive IA schemes are proposed. First, we propose an interference nulling based cognitive IA scheme. In such a scheme, both co-tier and cross-tier interferences are aligned into the orthogonal subspace at each IoT receiver, which means all the interference can be perfectly eliminated without causing any performance degradation on the macrocell. However, it is known that the interference nulling based IA algorithm achieves its optimum only in high signal to noise ratio (SNR) scenarios, where the noise power is negligible. Consequently, when the imposed interference-free constraint on the femtocell can be relaxed, we also present a partial cognitive IA scheme that further enhances the network performance under a low and intermediate SNR. Additionally, the feasibility conditions and capacity analyses of the proposed schemes are provided. Both theoretical and numerical results demonstrate that the proposed cognitive IA schemes outperform the traditional orthogonal precoding methods in terms of network capacity, while preserving for macrocell users the desired quality of service.