Deep Reinforcement Learning in Parameterized Action Space

Matthew Hausknecht, Peter Stone
11/13/2015
cs.AIcs.LGcs.MA

Official code: mhauskn/dqn-hfo (79 stars, MIT, last commit 2018-08-19).

Abstract

Recent work has shown that deep neural networks are capable of approximating both value functions and policies in reinforcement learning domains featuring continuous state and action spaces. However, to the best of our knowledge no previous work has succeeded at using deep neural networks in structured (parameterized) continuous action spaces. To fill this gap, this paper focuses on learning within the domain of simulated RoboCup soccer, which features a small set of discrete action types, each of which is parameterized with continuous variables. The best learned agent can score goals more reliably than the 2012 RoboCup champion agent. As such, this paper represents a successful extension of deep reinforcement learning to the class of parameterized action space MDPs.

View on arXivView PDF

Code Implementations(1)

mhauskn/dqn-hfoOfficial100%
MIT
7923C++, CMakeApr 20, 20158 years agoMIT

Similar Papers

Geometry of Neural Reinforcement Learning in Continuous State and Action Spaces

85% match2025

Multi-Pass Q-Networks for Deep Reinforcement Learning with Parameterised Action Spaces

85% match1 repo2402019

Action valuation of on- and off-ball soccer players based on multi-agent deep reinforcement learning

85% match2023

General Policy Evaluation and Improvement by Learning to Identify Few But Crucial States

84% match1 repo62022

Context-Sensitive Abstractions for Reinforcement Learning with Parameterized Actions

84% match2025

Action Branching Architectures for Deep Reinforcement Learning

84% match1 repo1222017

Cite this paper

@article{hausknecht2015deep,
  title  = {Deep Reinforcement Learning in Parameterized Action Space},
  author = {Matthew Hausknecht and Peter Stone},
  year   = {2015},
  eprint = {1511.04143},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1511.04143}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1511.04143.svg)](https://codeofpaper.com/paper/1511.04143)

Discussion