Multiple Access for Massive Machine Type Communications

Abstract

The internet we have known thus far has been an internet of people, as it has connected people with one another. However, these connections are forecasted to occupy only a minuscule of future communications. The internet of tomorrow is indeed: the internet of things. The Internet of Things (IoT) promises to improve all aspects of life by connecting everything to everything. An enormous amount of effort is being exerted to turn these visions into a reality. Sensors and actuators will communicate and operate in an automated fashion with no or minimal human intervention. In the current literature, these sensors and actuators are referred to as machines, and the communication amongst these machines is referred to as Machine to Machine (M2M) communication or Machine-Type Communication (MTC). As IoT requires a seamless mode of communication that is available anywhere and anytime, wireless communications will be one of the key enabling technologies for IoT. In existing wireless cellular networks, users with data to transmit first need to request channel access. All access requests are processed by a central unit that in return either grants or denies the access request. Once granted access, users' data transmissions are non-overlapping and interference free. However, as the number of IoT devices is forecasted to be in the order of hundreds of millions, if not billions, in the near future, the access channels of existing cellular networks are predicted to suffer from severe congestion and, thus, incur unpredictable latencies in the system. On the other hand, in random access, users with data to transmit will access the channel in an uncoordinated and probabilistic fashion, thus, requiring little or no signalling overhead. However, this reduction in overhead is at the expense of reliability and efficiency due to the interference caused by contending users. In most existing random access schemes, packets are lost when they experience interference from other packets transmitted over the same resources. Moreover, most existing random access schemes are best-effort schemes with almost no Quality of Service (QoS) guarantees. In this thesis, we investigate the performance of different random access schemes in different settings to resolve the problem of the massive access of IoT devices with diverse QoS guarantees. First, we take a step towards re-designing existing random access protocols such that they are more practical and more efficient. For many years, researchers have adopted the collision channel model in random access schemes: a collision is the event of two or more users transmitting over the same time-frequency resources. In the event of a collision, all the involved data is lost, and users need to retransmit their information. However, in practice, data can be recovered even in the presence of interference provided that the power of the signal is sufficiently larger than the power of the noise and the power of the interference. Based on this, we re-define the event of collision as the event of the interference power exceeding a pre-determined threshold. We propose a new analytical framework to compute the probability of packet recovery failure inspired by error control codes on graph. We optimize the random access parameters based on evolution strategies. Our results show a significant improvement in performance in terms of reliability and efficiency. Next, we focus on supporting the heterogeneous IoT applications and accommodating their diverse latency and reliability requirements in a unified access scheme. We propose a multi-stage approach where each group of applications transmits in different stages with different probabilities. We propose a new analytical framework to compute the probability of packet recovery failure for each group in each stage. We also optimize the random access parameters using evolution strategies. Our results show that our proposed scheme can outperform coordinated access schemes of existing cellular networks when the number of users is very large. Finally, we investigate random non-orthogonal multiple access schemes that are known to achieve a higher spectrum efficiency and are known to support higher loads. In our proposed scheme, user detection and channel estimation are carried out via pilot sequences that are transmitted simultaneously with the user's data. Here, a collision event is defined as the event of two or more users selecting the same pilot sequence. All collisions are regarded as interference to the remaining users. We first study the distribution of the interference power and derive its expression. Then, we use this expression to derive simple yet accurate analytical bounds on the throughput and outage probability of the proposed scheme. We consider both joint decoding as well as successive interference cancellation. We show that the proposed scheme is especially useful in the case of short packet transmissions