Learning from Outside the Viability Kernel: Why we Should Build Robots that can Fall with Grace

06/18/2018
by   Steve Heim, et al.
0

Despite impressive results using reinforcement learning to solve complex problems from scratch, in robotics this has still been largely limited to model-based learning with very informative reward functions. One of the major challenges is that the reward landscape often has large patches with no gradient, making it difficult to sample gradients effectively. We show here that the robot state-initialization can have a more important effect on the reward landscape than is generally expected. In particular, we show the counter-intuitive benefit of including initializations that are unviable, in other words initializing in states that are doomed to fail.

READ FULL TEXT

page 3

page 4

research
05/31/2016

Information Theoretically Aided Reinforcement Learning for Embodied Agents

Reinforcement learning for embodied agents is a challenging problem. The...
research
05/06/2020

Active Preference-Based Gaussian Process Regression for Reward Learning

Designing reward functions is a challenging problem in AI and robotics. ...
research
10/23/2022

Active Exploration for Robotic Manipulation

Robotic manipulation stands as a largely unsolved problem despite signif...
research
03/03/2023

Learning Stabilization Control from Observations by Learning Lyapunov-like Proxy Models

The deployment of Reinforcement Learning to robotics applications faces ...
research
11/17/2020

Curiosity Based Reinforcement Learning on Robot Manufacturing Cell

This paper introduces a novel combination of scheduling control on a fle...

Please sign up or login with your details

Forgot password? Click here to reset