Machine Learning And Edge Computing
Machine learning has recently found myriads of applications in communications and networks, involving almost all aspects from physical layer, MAC layer, network layer up to the application layer. As a key enabling technology of artificial intelligence (AI), machine learning approaches, including supervised learning, unsupervised learning and reinforcement learning, have the potential to greatly improve the communication network performance via empirical learning capabilities based on collected data from the dynamic communication environment.
Edge computing, on the other hand, is another inter-disciplinary module that intertwines with the optimal design of communications and networks, where the traditional communication nodes such as base stations, cell phones, UAVs, IoT devices, etc., are equipped with additional and different level of computing capabilities to handle their local computing tasks of different level of complexity. The computing tasks might include traditional baseband signal processing for communication links, traditional machine learning tasks like image or video feature extraction, or other more advance tasks in social networks, cyber-physical security, etc. The additional computing capability at the edge inspires new design trade-offs between communication and computing, both of which might benefit from the up-surging paradigm of machine learning enabled design.
In this field of research, we aim to apply machine learning methods to critical and challenging communication and edge-computing problems, and design optimization and decision-making algorithms with tailored consideration for the specific communication and edge-computing context, such that the traditionally difficult, dynamic, and intractable problems can be addressed intelligently with hopefully low complexity and high efficiency. Specific application fields include UAV communications, UAC networks, cross-layer and large scale network optimization, on-board intelligence and joint communication design, computation offloading for IoT devices, among others.