EconCS Seminar | Sarah Keren (Technion) | Encouraging Automated Agents to Behave Nicely 

Abstract: 

Autonomous AI agents are deployed in increasingly complex environments where they must account for the presence of other agents while trying to achieve their objectives.  Moreover, such agents may require assistance from other agents to efficiently accomplish their assigned task, or even be able to complete it. 

The overarching objective of this work is to provide theoretical and computational foundations that will allow agents to learn and adopt cooperative and collaborative behaviors autonomously. As a first step for accomplishing this objective, we need to equip agents with the ability to produce ad-hoc assessments of their ability to be helpful to other agents and of the potential benefit of assistive actions that may be performed by other agents.  

For this purpose, the Value of Assistance (VOA) will be presented, capturing the expected improvement in the performance of assistive actions. Computing VOA in different multi-agent and multi-robot settings will also be discussed. As an example, VOA estimation will be demonstrated for selecting which agent from a team of navigating robotic agents should receive localization information from a drone. Estimation of VOA will also be demonstrated for grasping and integrated task and motion planning settings. 

 

Location: 

Seminar room, Feldman Building, Edmond Safra Campus