Multi-Agent Competitive Reinforcement Learning
2.7 هزار بار بازدید -
4 سال پیش
-
This video corresponds to our
This video corresponds to our paper, Natural Emergence of Heterogeneous Strategies in Artificially Intelligent Competitive Teams, to be presented in Robotics: Science and Systems (RSS), 2020 Workshop on Heterogeneous Multi-Robot Task Allocation and Coordination.
We propose to use Reinforcement Learning with Graph Neural Networks in a Guards vs Attackers scenario called FortAttack, which allows us to scale to a large number of agents and remain sample efficient. We observe that multi agent competition naturally leads to the emergence of complex strategies, which require strong coordination within a team. We also observe the emergence of heterogeneous behavior amongst homogeneous agents when such behavior can lead to the team's success
Code: https://github.com/Ankur-Deka/Emergen...
Workshop: https://sites.google.com/view/mrta-rs...
We propose to use Reinforcement Learning with Graph Neural Networks in a Guards vs Attackers scenario called FortAttack, which allows us to scale to a large number of agents and remain sample efficient. We observe that multi agent competition naturally leads to the emergence of complex strategies, which require strong coordination within a team. We also observe the emergence of heterogeneous behavior amongst homogeneous agents when such behavior can lead to the team's success
Code: https://github.com/Ankur-Deka/Emergen...
Workshop: https://sites.google.com/view/mrta-rs...
4 سال پیش
در تاریخ 1399/04/14 منتشر شده
است.
2,773
بـار بازدید شده