Next-generation networks such as the Internet of Things (IoT), connected and autonomous vehicles (CAVs), social networks, augmented and virtual reality (AR/VR), as well as other wireless connected systems are expected to have enhanced capacity to autonomously process data and operate. However, such intelligent systems still need to work under human intervention, and act fast enough to the irrational and psychological behaviors. The goal of this research direction is to introduce novel networking solutions and enable intelligent devices quickly interpret and response to complex human behavior.
In particular, the basic idea of such networking solutions is to model human behavioral patterns and choose the most proper corresponding networking operations. However, human behavioral patterns can be affected by multiple factors, including human psychology, expertise, personal habits, and so on, which makes it hard to be captured and formulated. Thus, a key goal here is to develop novel inverse learning solutions for effective interpretation of complex human behavior.
This includes the following tasks:
1) Development of inverse learning algorithms to detect selfish human behavior in cooperative multi-agent solutions. We will propose inverse reinforcement learning solutions to interpret human strategies from limited observation of human operations. Based on such interpretation, we can defend the multi-agent systems from selfish behaviors.
2) Development of meta inverse learning solutions that not only can perform well in expected tasks and even in the unseen tasks, with reduced training data, as well as computation complexity.
3) Application of the developed solutions in emerging and future 6G services such as AR/VR, holography, CAVs, and digital twins.