Intelligent Communication Networks and Internet of Things
The symbiosis between computing and communication means that current communication network architectures include computational elements, making them systems that process and transmit data. Another striking feature of communication networks is the incorporation of intelligence at the edge, in the access network, and in the network's core, which goes far beyond the execution of machine learning algorithms with data generated by users and sensors. Although the use of AI in communication networks is at an early stage, it will be a fundamental element in future 6G networks.
The INCT scientific project defines thematic lines to face scientific, technological, and innovation challenges to achieve the proposed objectives. The INCT's lines of research are: the Internet of Things, AI, edge computing, 5G/6G networks, optical access networks, network virtualization, vehicular networks, security, smart cities, digital health, industry, and environmental preservation.
The large number of devices connected to the Internet of Things requires the handling of a high volume of data generated by thousands of sensors, requiring solutions that meet scalability, geographic distribution, mobility, heterogeneity, security, and privacy requirements. The heterogeneity of devices, intermittent connections, and the diversity of data structures generated by sensors are some of the technological constraints that hinder the implementation of a pervasive IoT. Other factors of capital importance to be observed are data locality-oriented computing and the high volume of data to be transmitted in the network. Adaptive allocation and resource orchestration are challenges to overcome in large-scale IoT networks with thousands of sensors.
Currently, the execution of machine learning algorithms is typically batch, offline, and centralized. Network management and its services require massive execution of distributed data in real-time. In many situations, the temporal validity of the generated data is limited, requiring the reduction of latency in communication and processing. Furthermore, data transmission in a distributed environment is subject to the quality of communication channels, network congestion, and the energy available on mobile devices. An additional substantial restriction arises from adopting the new General Data Protection Law - LGPD. The federated learning technique addresses the data privacy restriction. However, there are numerous challenges in federated learning, such as characterizing the eligibility of clients to participate in the learning and the asynchrony in data transmissions.
Edge computing consists of bringing services offered by computing clouds to the edge of the network, mitigating the long delays to access servers in computing clouds. Edge devices are typically heterogeneous and have limited resources. Furthermore, there still need to be comprehensive models and mature technologies for intelligent orchestration and dynamic adjustment of physical or virtualized resources to meet QoS requirements under varying network and workload conditions. The generalization of edge-to-cloud processing includes devices located in the access network and in the core, in a resource continuum between edge and cloud. Resource allocation in this continuum requires complex and adaptive solutions, which are possible only through AI. Furthermore, the execution of AI algorithms at the edge of the network, the so-called intelligence at the edge, brings several challenges in reducing the need for data communication and energy consumption.
The recent 5G cellular network technology will enable many applications with strict latency requirements and high bandwidth demands. Adopting network virtualization and edge computing as fundamental elements of 5G (native) networks allows the rapid creation of customized networks for different sectors of the Economy (verticals). However, service automation depends on the ability to dynamically reconfigure the virtualized network. Such capability requires the consideration of numerous parameters, which is only possible through AI-based solutions. 6G technology, the successor to 5G, will deal with quality requirements that are orders of magnitude different from those of 5G networks, which requires the adoption of AI as a native technology in the architecture of 6G networks.
The advent of the Internet of Things, cloud environments (edge), and 5G networks challenge current network and system security solutions. The Internet of Things and 5G technology enable a wide range of heterogeneous and resource-constrained devices to connect to the Internet, forcing new behaviors in data traffic.
