This paper introduces a novel resource allocation framework for the coexistence of enhanced mobile broadband (eMBB) and ultra-reliable low latency communication (URLLC) services in a non-orthogonal multiple access (NOMA) system. The proposed solution utilizes a two-layer deep reinforcement learning (DRL) algorithm to optimize the dynamic allocation of subbands and power for both URLLC and eMBB users. In the top layer, a deep Q-network (DQN) is employed for discrete subband allocation while the bottom layer uses a multi-agent DRL approach to refine the continuous power allocation. This hierarchical approach ensures that low latency and reliability requirements of the URRLC services are met, without compromising the high throughput demands of eMBB. Extensive simulations validate the efficiency of the proposed algorithm, demonstrating improved system performance in terms of power and subband optimization for both user types.

International audience

The Internet of Things (IoT) is a major driving use case of future communication systems. 5G and beyond networks should be properly designed to meet the quality of service (QoS) requirements of IoT devices in terms of massive access, data rates, latency and reliability. In this paper, we study the uncoordinated spectrum and power allocation problems in an uplink IoT network. In the considered setting, IoT devices aim at organizing their transmissions, without any coordination, on the available frequency channels and power levels. To enhance performance, we propose the deployment of full-duplex relays as well as the use of non-orthogonal multiple access and queuing at the level of the relays. A new algorithm, based on the multi-player multi-armed bandits framework, is proposed with the aim of reducing the transmit power of IoT devices. Simulation results validate the convergence of the proposed algorithm. Results also confirm the superior performance of the proposed technique in terms of satisfying QoS requirements and reducing the needed user transmit power when compared to two baseline schemes.

To achieve better connectivity in future communication networks, the deployment of different types of access points (APs) is underway. APs are expected to be equipped with self-organizing capabilities to reduce costs. Moreover, due to the spectrum crunch, frequency reuse among the deployed APs is inevitable, exacerbating the problem of inter-cell interference (ICI). Therefore, ICI mitigation in self-organizing networks (SONs) is commonly identified as a key radio resource management mechanism to enhance performance. To this end, this paper proposes a novel solution for the uncoordinated channel and power allocation problems. Based on the multi-armed bandits (MAB) framework, the proposed technique does not require any communication between the APs. The case of varying channel rewards across APs is considered. In contrast to previous work on channel allocation using the MAB framework, APs are permitted to choose multiple channels for transmission. Moreover, non-orthogonal multiple access is used, allowing multiple APs to access each channel simultaneously. This results in an MAB model with varying channel rewards, multiple plays and non-zero reward on collision. The proposed algorithm has an expected regret in the order ofO(log2T), with extensive numerical results revealing it significantly outperforms a well-known baseline algorithm in terms of energy efficiency.

In this paper, an algorithm based on the multiplayer multi-armed bandit (MAB) framework is proposed to solve an uncoordinated spectrum access problem. The proposed technique does not require any communication or coordination between users. The case of varying channel rewards across users is considered. In contrast to previous work, the users are permitted to choose multiple channels for transmission, resulting in a MAB model with multiple plays. The proposed algorithm has an expected regret of the order O(log2T), which is validated by simulation results.

In this paper, we study the traffic-aware resource allocation problem for a system with mixed traffic types. The considered framework encompasses real-time (RT) users having strict QoS requirements (in terms of amount of data and latency), and best-effort (BE) users for which the system tries to strike a balance between throughput and fairness. The resource allocation problem is studied in different contexts: orthogonal and non-orthogonal multiple access (OMA and NOMA respectively) in either centralized or distributed antenna systems (CAS and DAS respectively). Following the formulation of the resource optimization problem, we propose a low complexity suboptimal solution based on matching theory for each system context. We also propose an iterative approach to determine the number of subbands per antenna for the DAS contexts. The proposed techniques aim at guaranteeing the requirements of RT users while maximizing the utility function of BE users. Simulation results show that the proposed allocation method based on matching theory greatly outperforms a previously proposed greedy approach, especially in terms of RT users satisfaction.

Unmanned aerial vehicle (UAV)-base stations (BS) are envisioned to provide wireless access to areas where the existing wireless infrastructure is either not deployed, damaged or simply congested. In this paper, the problem of minimizing the transmit power of a UAV-BS, while serving users with their QoS requirements and accounting for the wireless backhaul limitation of the UAV, is investigated. To this end, an algorithm that finds the optimal bandwidth assignment in the backhaul link, the optimal 3D position of the UAV, as well as the transmit power distribution in the access and the backhaul links is proposed. Simulation results show that, using the proposed approach, the needed power is greatly reduced, when compared to a strategy that separates between the frequency band used in the access and the backhaul links. In addition, the performance enhancement of a UAV-enabled system over a traditional one in which the macro base station (MBS) directly serves users is shown.

This paper proposes a new traffic-aware resource allocation technique, employing non-orthogonal multiple access (NOMA) in a downlink distributed antenna system (DAS). The studied framework consists of users with mixed traffic types: real-time (RT) users having strict QoS requirements (in terms of amount of data and latency), and best-effort (BE) users aiming to maximize their throughput and fairness. After formulating the resource and power optimization problem, we propose a low complexity sub-optimal algorithm that aims at guaranteeing the requirements of RT users while maximizing the utility function of BE users. Simulation results show a remarkable performance enhancement of the proposed algorithm over baseline techniques in terms of RT users satisfaction. Also, the proposed technique achieves near optimal fairness for BE users while maximizing their average throughput.

Non-Orthogonal Multiple Access (NOMA) has proven itself as a serious candidate for multiple access schemes in the 5thgeneration of wireless communication systems. In this paper, we propose new resource allocation techniques that allow the coupling of NOMA with multiple-input-multiple-output (MIMO) systems. As in most of downlink NOMA systems, the proportional fairness (PF) scheduler is used. As for power allocation (PA), most previous works assume an equal PA between antennas and subbands, which is sub-optimal. In this paper, we first propose a technique to reduce the complexity of the PF scheduler, and then we introduce an iterative waterfilling-based power allocation to the Single-User MIMO (SU-MIMO) case. Extensive simulations show that the novel PA technique improves the quality of experience of users suffering from bad channel conditions, as well as the user fairness.