GDC is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them. Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

Use the GDC 2021 Session Viewer to browse and sort the ever-growing list of sessions by pass type, topic, and format. You will be able to build your personal schedule directly within the event platform, once live early July 2021.

Machine Learning Summit: Creating Cooperative Character Behaviors Using Deep Reinforcement Learning

Marwan Mattar  (Senior Manager of Machine Learning, Unity Technologies)

Vincent-Pierre Berges  (Senior Machine Learning Engineer, Unity Technologies)

Pass Type: All Access Pass, Summits Pass

Topic: Programming

Format: Session

Vault Recording: TBD

Audience Level: N/A: I prefer to present live on-site but I am open to delivering a virtual presentation

Creating cooperative character behaviors is very challenging and is costly to studios of all sizes. Deep reinforcement learning can be a great approach because a developer can articulate the desired objectives or goals and have the machines learn the best behavior. This can impact the economics of the development of the game. However, most DRL algorithms are intended to solve single character tasks (hence no cooperation between characters). However, there are new advancements in cooperative DRL tools, such as centralized critic, that can be leveraged to create cooperative character behaviors in a game. In this presentation, we will go over these topics and illustrate a real-life example of a studio using deep reinforcement learning to create cooperative characters in their game.


After attending this presentation, the audience member will understand a general framework for creating cooperative character behaviors using deep reinforcement learning.

Intended Audience

The session is intended for game AI engineers and designers but is open to any game developers who is implementing deep reinforcement learning in their games. No prerequisites necessary as we will provide a high level overview of deep reinforcement learning.