Log In Sign Up

Deep Latent Competition: Learning to Race Using Visual Control Policies in Latent Space

by   Wilko Schwarting, et al.

Learning competitive behaviors in multi-agent settings such as racing requires long-term reasoning about potential adversarial interactions. This paper presents Deep Latent Competition (DLC), a novel reinforcement learning algorithm that learns competitive visual control policies through self-play in imagination. The DLC agent imagines multi-agent interaction sequences in the compact latent space of a learned world model that combines a joint transition function with opponent viewpoint prediction. Imagined self-play reduces costly sample generation in the real world, while the latent representation enables planning to scale gracefully with observation dimensionality. We demonstrate the effectiveness of our algorithm in learning competitive behaviors on a novel multi-agent racing benchmark that requires planning from image observations. Code and videos available at


page 2

page 3

page 4

page 5

page 6

page 9

page 10

page 13


Options as responses: Grounding behavioural hierarchies in multi-agent RL

We propose a novel hierarchical agent architecture for multi-agent reinf...

Emergent Complexity via Multi-Agent Competition

Reinforcement learning algorithms can train agents that solve problems i...

Influencing Towards Stable Multi-Agent Interactions

Learning in multi-agent environments is difficult due to the non-station...

Contingencies from Observations: Tractable Contingency Planning with Learned Behavior Models

Humans have a remarkable ability to make decisions by accurately reasoni...

Dream to Control: Learning Behaviors by Latent Imagination

Learned world models summarize an agent's experience to facilitate learn...

TLeague: A Framework for Competitive Self-Play based Distributed Multi-Agent Reinforcement Learning

Competitive Self-Play (CSP) based Multi-Agent Reinforcement Learning (MA...

Investigating Taxi and Uber competition in New York City: Multi-agent modeling by reinforcement-learning

The taxi business has been overly regulated for many decades. Regulation...