The art of AI & Signals in communication systems
Machine learning will have a substantial impact on telecommunication in upcoming years, where problems spanning all scales of the universe, from the identification of signals to the detection of anomalies in data transmission and the prediction of malicious events, and will see critical improvements due to applications of deep learning methods. These advances are made possible through the permeation of recent machine learning advances in the ICT community, creating unique insights through the combination of machine learning expertise and communication domain knowledge.
Machine learning has seen a surge of interest in communication technologies, with techniques ranging from computer vision, natural language processing, generative modelling, and reinforcement learning leading to new methods to solve long-standing problems in the classification, simulation and analysis of information transmitting systems.
As modern communications systems increase in both quantity and complexity, the need for sophisticated, intelligent algorithms in the signal characterization space concurrently rises. Designing and optimizing application-specific artificial intelligence (Ai) networks for use on time domain data can increase detection sensitivity and identification speed while minimizing inherent latency in conventional digital signal processing (DSP) techniques. Applied Ai can eliminate the need for common data preprocessing techniques and make decisions on raw digitized data. From electronic warfare (EW) to spectrum management,YOTAVIS is solving complex problems associated with signal characterization,bringing appliedAi to the edge and changing the way communications systems operate.
Artificial Intelligence (AI), well known from computer science disciplines, are beginning to emerge in the wireless communications and have recently received much attention as a key enabler for future 5G and beyond wireless networks. These AI approaches including Machine Learning (ML), Deep Learning (DL) and Deep Reinforcement Learning (DRL) approaches have been gradually applied to wireless communication systems for various purposes which extensively improve the performance of wireless communication systems and users’ QoE. Therefore, AI technologies have a great potential to meet the various requirements of seamless wide-area coverage, low-power massive-connections, low latency high-reliability, and many other scenarios.
Due to the new features of future communications, such as complex scenarios with unknown channel models, high speed and accurate processing requirements, traditional methods are no longer suitable which brings much more potential application of AI. Just as DL technology has become a new hotspot in the research of physical-layer wireless communications and challenges conventional communication theories.
Currently DL-based methods show promising performance improvements but lack of solid analytical tools and universal network architectures. In addition to the traditional neural network-based data-driven model, the model-driven deep network model and the DRL model (i.e. DQN) which combined DL with reinforcement learning are more suitable for dealing with future communication systems which can be modelled with interpretability. Moreover, most of current studies focus on solving old problems such as estimation accuracy and resource allocation optimization in wireless communication systems. However, it is important to distinguish new capabilities created by AI technologies and rethink wireless communication systems based on AI-driven schemes. Therefore, the old theory will be supplemented and updated to a large extent when solving the old problems with the new method of AI. At the same time, the problems brought by the introduction of AI technology into communication, such as how to reduce the complexity of AI algorithm to make it suitable for lightweight devices and so on are also important directions in the future.