Identifying Temporal Patterns in Multi-Agent Systems
The Smile-IT project aims to develop a multi-agent framework for studying and managing modern distributed networked systems (telecom networks, smart grids, traffic networks…) that contain a large number of entities or agents, both machine and human, which strive to achieve their personal objectives. The framework developed within the project will guide these entities, either through direct control or by way of incentives, in order to achieve system-wide optimal behaviour, satisfy global objectives and adhere to the system’s operational constraints in the face of diverging and incompatible personal goals.
A key aspect of the project is the identification of temporal patterns. Networked systems often exhibit temporal patterns, e.g rush hour, day/night cycles , weekend/weekday, ... The multi-agent system should be able to learn and identify these temporal patterns in the system behavior. These learned patterns could then be used to respond faster to recurring events (e.g. rush hour, day/night cycles in energy usage) which improves the robustness of the system. This also allows the system to be proactive rather than always re-adapting to a changing environment and continuously re-learning the same behaviours.
Various methods exist to cluster states into single events and model temporal connections between them. These methods are mainly Bayesian inference methods, which maintain probabilistic models over (and transitions between) states, such as Hidden Markov Models or Gaussian Processes, and compute the probability of observing a specific temporal pattern.
In this thesis you will study methods to detect and identify recurring patterns in the behavior of network systems. You will create a small case study of a traffic simulation (see http://www.cs.utexas.edu/~aim/) in which the traffic patterns vary regularly over time. You will then evaluate methods to detect these patterns and identify the current phase the system is in, based on the current state of the system (e.g. waiting time at the different intersections).
- SMILE-IT: https://ai.vub.ac.be/smileit/
- Game Theory and Multi-agent Reinforcement Learning