Matin Kheirkhahan - Artificial Intelligence M.S. graduate
Mohammadreza Kangavari - Assistant Professor
Farimah Farahmandi

As autonomous agents proliferate in the real world, both in software and robotic settings, they will increasingly need to band together for cooperative activities with previouslyunfamiliar teammates. Having autonomous agents capable of learning its previously unknown teammates’ preferences and acting in harmony with them to achieve the system’s end has been declared as a challenge to the AI. An agent capable of adhoc teamwork is one that can effectively cooperate with multiplepotential teammates on a set of collaborative tasks. This agent has to learn other agents’ behaviours and preferences in orderto be able to predict their actions on some level. This paper presents an empirical study of ad hoc teamwork, takingadvantage of Observer System’s merits. Specifically, we evaluate a learning mechanism for on-line behaviour generation on the part of a single ad hoc team agent that must collaborate with other unknown teammates in the pursuit domain.

Ad hoc teams, Cooperation, Multiagent System, Observer Agent, Learning