Machine Learning and AI are powerful new tools that promise a technical revolution. SATORI – SmArt neTwOrking in the eRa of aI – represents a joint collaboration between Trinity College Dublin and Tsinghua University, with the aim of applying these techniques to the next generation of communication networks, as well as smart cities. The research efforts of the two groups follow three separate tracts, representing three areas in the field of communication networks that can benefit from the application of these tools. These tracts are:
1. AI for Communications in Smart Cities
By focusing on characterizing the mobile traffic, web and information usage traces based on large-scale and long-duration spatial-temporal big data, which is collected from a commercial mobile operator with more than 10 thousand base stations and 6.5 million users spanning over some months in large scale of city, this project aims to qualitatively visualize and quantitatively characterize the spatio-temporal human mobility behaviors in the physical-cyber space in terms of mobility, traffic consumption, social activity, etc. Based on these fundamental findings and credible models, this project plan to further investigate how to utilize these important insights to deal with problems encountered with the current urban mobile networks, urban management, and robust cyber-physical systems.
2. AI for Communications in Moving Networks
In moving networks, like V2V communications, not only the model of the transmission environment is hard to describe, but also the environment as well as the mission of the communication tasks are changing over time, based on, for example, the context information. This calls for learning-based approaches to be capable of model-free optimization, and reinforcement learning to track the changing contexts. In addition, to coordinate the distributed devices in the moving network, multi-agent learning techniques are a promising solution, which also needs to be carefully tuned for the balance of complexity and performance.
3. AI for Communications in Shared Networks
Sharing in networks covers the range of technologies and scenarios that allow heterogeneous resources (or resources of heterogeneous ownership) to be accessed and utilised in new ways. This introduces new degrees of freedom that may be translated to gains in network quality, such as efficiency or reliability, defined individually for each actor. This increased complexity, along with heterogeneous, distributed, and often dynamic context of shared networks, makes Machine Learning a promising solution to support their design, management, and operation.
The project is co-funded by Science Foundation Ireland and the National Natural Science Foundation of China.