Institute Homepage
Institute Homepage EN Sign In

Back

Autonomous Learning Embodied Vision Conference Paper 2020

Sample-efficient Cross-Entropy Method for Real-time Planning

Paper Code Spotlight-Video
Autonomous Learning
Cristina Pinneri
Robust Machine Learning
Sebastian Blaes
Postdoctoral Researcher
Embodied Vision
Jan Achterhold
Embodied Vision
Jörg Stückler
Max Planck Research Group Leader
Autonomous Learning
Michal Rolinek
Empirische Inferenz, Autonomous Learning
Georg Martius
Senior Research Scientist

Trajectory optimizers for model-based reinforcement learning, such as the Cross-Entropy Method (CEM), can yield compelling results even in high-dimensional control tasks and sparse-reward environments. However, their sampling inefficiency prevents them from being used for real-time planning and control. We propose an improved version of the CEM algorithm for fast planning, with novel additions including temporally-correlated actions and memory, requiring 2.7-22x less samples and yielding a performance increase of 1.2-10x in high-dimensional control problems.

Author(s): Cristina Pinneri and Shambhuraj Sawant and Sebastian Blaes and Jan Achterhold and Joerg Stueckler and Michal Rolinek and Georg Martius
Links:
Book Title: Conference on Robot Learning 2020
Year: 2020
Project(s):
BibTeX Type: Conference Paper (inproceedings)
State: Published
URL: https://corlconf.github.io/corl2020/paper_217/
Electronic Archiving: grant_archive

BibTeX 

@inproceedings{PinneriEtAl2020:iCEM,
  title = {Sample-efficient Cross-Entropy Method for Real-time Planning},
  booktitle = {Conference on Robot Learning 2020},
  abstract = {Trajectory optimizers for model-based reinforcement learning, such as the Cross-Entropy Method (CEM), can yield compelling results even in high-dimensional control tasks and sparse-reward environments. However, their sampling inefficiency prevents them from being used for real-time planning and control. We propose an improved version of the CEM algorithm for fast planning, with novel additions including temporally-correlated actions and memory, requiring 2.7-22x less samples and yielding a performance increase of 1.2-10x in high-dimensional control problems.},
  year = {2020},
  author = {Pinneri, Cristina and Sawant, Shambhuraj and Blaes, Sebastian and Achterhold, Jan and Stueckler, Joerg and Rolinek, Michal and Martius, Georg},
  url = {https://corlconf.github.io/corl2020/paper_217/ }
}