The application of Artificial Intelligence is affecting the design and the application of the wireless networking. In the Artificial Intelligence techniques for Wireless Multimedia Networking (AI4WMN) project we aim at exploiting AI in two emerging fields of the wireless networking: namely the Internet of Things security and the Multimedia Unmanned networks.
In these two representative use cases the AI can be used to adapt algorithms to the main features of the environments where these algorithms have to be applied. Both the data and the scenarios where these algorithms have to be run are dymanic and have specific constraints. The AI4WMN project encompasses all these aspects by providing solutions that, by leveraging the AI, improves the system behavior and its performance.
We will provide the use of AI for the IoT security where challenges related to the tradeoff that we have in these systems due to the huge amount of data on a side and the simplicity of the involved devices of the other side is in place. Intrusion and anomaly detection methods will be defined and the AI using reinforced learning approaches will be used.
The second field is when the wireless networks involve Unmanned devices (e.g., Unmanned Aerial Vehicles, UAV) using the IoT and the network for operating multimedia applications. This proposal covers several key aspects of this kind of communications: from the design of a scalable communication architecture, to the development of new network security algorithms and multimedia service provisioning, to the implementation of simulation frameworks that adapt to time and space-wireless conditions.
The research in AI4WMN will be innovative under different perspectives.
- As for the IoT we will address mainly its security issues and the use of AI; we will demonstrate how AI can be tailored to match the requirements of IoT security still respecting the constraints that an IoT architecture has.
- As for the Augmented Reality we will show the advantages of advanced coding techniques for Mixed and Augmented Reality video encoding and mobile streaming, as well as MAR visual data processing by devising filtering.
- As for the Multi Agent Reinforcement Learning (MARL) we will demonstrate how temporal logics to specify multi-agent goals and hierarchical representations can improve scalability in the complex AI4WMN use cases.