Tel-Aviv University - Computer Science Colloquium
Thursday, February 25, 13:15-14:15 (Note unusual time)
at 13:00 My research contributes a multi-agent system organization and a new learning approach for teams of autonomous agents acting in real-time, noisy, collaborative, and adversarial environments.
First, I will present a flexible team structure which allows agents to decompose the task space into roles and formations. Team organization is achieved by the introduction of a "locker-room agreement" as a collection of shared conventions, including team plans and pre-defined conditions for agents to switch roles in coordination.
Second, I will introduce layered learning, a general-purpose machine learning paradigm for complex domains in which learning a mapping directly from agents' sensors to their actuators is intractable. Given a hierarchical task decomposition, layered learning allows for learning at every level of the hierarchy, with learning at each level directly affecting learning at the next higher level.
Third, I will introduce a new multi-agent reinforcement learning algorithm, namely team-partitioned, opaque-transition reinforcement learning (TPOT-RL). TPOT-RL is designed for domains in which agents cannot necessarily observe the state changes when other team members act.
All three main contributions of my work are combined within a fully functioning multi-agent system that incorporates learning in a real-time, noisy domain with teammates and adversaries, namely simulated robotic soccer domain. Empirical results validate all three contributions in controlled experiments. The generality of the contributions is verified by applying them to the real robotic soccer and network routing domains.
The talk will also report on results in the RoboCup international competitions. The contributions of my work were fully incorporated in the CMUnited-98 simulator RoboCup-98 champion team and partly in the CMUnited-97 small-robot RoboCup-97 champion team.
For colloquium schedule, see http://www.math.tau.ac.il/~matias/colloq.html