Emergent tool use from multi-agent autocurricula
This paper explores the emergent behaviors in multi-agent environments, specifically focusing on dynamics like hide-and-seek, highlighting how complex behaviors can naturally arise from simple rules and interactions in a structured learning framework. © OpenAI
Control Strategies for Physically Simulated Characters Performing Two-player Competitive Sports
This paper presents a learning framework using deep reinforcement learning to develop control policies for physically simulated athletes in sports like boxing and fencing. © Meta Research
A Hybrid Material Point Method for Frictional Contact with Diverse Materials
This paper provides an improved approach in simulating elastic objects, such as hair, rubber and skin, as well as their two-way coupled interactions with other materials. © ACM