Authors: Cristina Pinneri, Shambhuraj Sawant, Sebastian Blaes, Jan Achterhold, Joerg Stueckler, Michal Rolínek, Georg Martius
Cross-Entropy method is a powerful optimizer for model-based RL. But it is too slow for real time deployment. We introduce several adjustments and tricks to bridge the “real time” gap. The resulting iCEM is FAAAAST!
[1/6] How many random planning trajectories (30 steps ahead) do I need to get this behavior?— Georg Martius (@GMartius) August 19, 2020
See our newest paper on making CEM efficient: https://t.co/2XKmF5bQSw
More videos: https://t.co/Wo1oyeuvob#Robotics, #reinforcementlearning, #CEM pic.twitter.com/eOohCv3TDc