Internal Rewards Mitigate Agent Boundedness

By Jonathan Sorg, Satinder Singh, Richard L. Lewis. 2010

61 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Learning from Physical Human Corrections, One Feature at a Time (2018, background): “Ideally, the human’s correction will only affect the cup’s distance from the table; in practice, however, human actions are noisy and imperfect [7, 19, 20, 24], especially when kinesthetically maneuvering robotic manipulators while trying to carefully orchestrate their multiple degrees of freedom [1”
Emerging social awareness: Exploring intrinsic motivation in multiagent learning (2011, background): “Also, as argued in [12], [13], the mechanism of reward optimization is fundamentally different from reward-shaping [14], as the latter approach does not modify the optimal policy for a certain environment.”
Safe Reinforcement Learning (2015, background): “Researchers have shown that often the most obvious reward function for a problem is not the most appropriate (Singh et al., 2009, Sorg et al., 2010).”
Intrinsic Motivation and Reinforcement Learning (2013, background): “Sorg, Singh, and Lewis (2010) provide computational results to support the view that a key role played by reward functions is to attenuate the negative consequences of various types of agent limitations, such as lack of information, lack of adequate time to learn, or lack of efficient learning…”
A Review of Robot Learning for Manipulation: Challenges, Representations, and Algorithms (2019, background): “More generally, modified reward functions can help to overcome multiple forms of agent boundedness [352] and can be found via gradient descent [351], genetic programming [280], or other evolutionary methods [161].”
Reward Design via Online Gradient Ascent (2010, background): “The optimal reward parameters are given by the solution to the optimal reward problem [16, 17, 18]:”
Intrinsically Motivated Reinforcement Learning: An Evolutionary Perspective (2010, background): “(see [44] for further exploration of such compensation).”
Deep Learning for Reward Design to Improve Monte Carlo Tree Search in ATARI Games (2016, methodology): “Our work builds on PGRD (policy-gradient for reward-design; Sorg et al. 2010a), a method for learning reward-bonus functions for use in planning.”