Recent Advances in Multi-Agent Systems: decentralised coordination for situated agents

Situated agents are computational components embedded in the environments in which they operate (e.g., robotic platforms, sensors, etc.). Situated agents are crucial elements in many emerging applications such as environmental sensing, wide area monitoring and energy management to name a few. In such applications, agents must establish collaborative interactions so to achieve their individual and collective goals. Within this setting, a key problem is for agents to select individual actions so to optimize a system wide objective function. This problem is typically referred to as decentralised coordination.

This talk discusses research problems related to decentralised coordination presenting recent advances in this field. In particular, the talk focuses on the use of optimization techniques based on graphical models and constraint processing to solve the decentralised coordination problem. We discuss how iterative optimization algorithms (such as max-sum) can be used within this context and present innovative approximate solution techniques discussing their merits within our reference application scenario.
Finally, we conclude the talk by highlighting open problems and possible future venues of research within this field.