• Publications
  • Influence
Benchmarking Model-Based Reinforcement Learning
TLDR
We benchmark 11 MBRL algorithms and 4 MFRL algorithms across 18 environments based on the standard OpenAI Gym, including noisy environments. Expand
  • 119
  • 12
  • PDF
NerveNet: Learning Structured Policy with Graph Neural Networks
TLDR
In this work, we propose NerveNet to explicitly model the structure of an agent, which naturally takes the form of a graph and then predict actions for different parts of the agent. Expand
  • 105
  • 11
  • PDF
VirtualHome: Simulating Household Activities Via Programs
TLDR
We introduce the VirtualHome simulator that allows us to create a large activity video dataset with rich ground-truth by using programs to drive an agent in a synthetic world. Expand
  • 58
  • 6
  • PDF
Exploring Model-based Planning with Policy Networks
TLDR
We propose a novel MBRL algorithm, model-based policy planning (POPLIN), that combines policy networks with online planning. Expand
  • 57
  • 6
  • PDF
Learning to Generate Diverse Dance Motions with Transformer
TLDR
In this work, we introduce a novel system that can synthesize diverse dance motions by learning from a large-scale dataset with a comprehensive set of highly diverse dance movements. Expand
  • 3
  • 2
  • PDF
Access points selection in super WiFi network powered by solar energy harvesting
TLDR
In this paper, we study the access point selection strategy from users' perspectives and consider Super WiFi networks powered by solar energy harvesting. Expand
  • 11
  • 1
  • PDF
Neural Graph Evolution: Towards Efficient Automatic Robot Design
TLDR
We propose Neural Graph Evolution (NGE), which performs selection on current candidates and evolves new ones iteratively. Expand
  • 9
  • 1
  • PDF
Access Strategy in Super WiFi Network Powered by Solar Energy Harvesting: A POMDP Method
TLDR
We propose a practical and efficient framework for multiple base stations access strategy in an EH powered Super Wi-Fi network. Expand
  • 2
  • PDF
UniCon: Universal Neural Controller For Physics-based Character Motion
TLDR
In this paper, we propose a physics-based universal neural controller (UniCon) that learns to master thousands of motions with different styles by learning on large-scale motion datasets. Expand