Abstract
This chapter presents a brief summary of the texplore algorithm before fully describing and presenting the real time RL architecture. First, I present a typical example of a sequential model-based RL architecture. Then I present details on using Monte Carlo Tree Search for planning, including a description of the modified version of the uct algorithm (Kocsis and Szepesvári, 2006) that we use for planning. In Section 3.2, I present the parallel architecture for real time action, which puts model learning, planning, and acting on three parallel threads, such that actions can be taken as fast as required without being constrained by how long model updates or planning take. Finally, I summarize the chapter in Section 3.3.
This chapter contains material from two publications: (Hester et al., 2012; Hester and Stone, 2012b).
Access provided by Autonomous University of Puebla. Download to read the full chapter text
Chapter PDF
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2013 Springer International Publishing Switzerland
About this chapter
Cite this chapter
Hester, T. (2013). Real Time Architecture. In: TEXPLORE: Temporal Difference Reinforcement Learning for Robots and Time-Constrained Domains. Studies in Computational Intelligence, vol 503. Springer, Heidelberg. https://doi.org/10.1007/978-3-319-01168-4_3
Download citation
DOI: https://doi.org/10.1007/978-3-319-01168-4_3
Publisher Name: Springer, Heidelberg
Print ISBN: 978-3-319-01167-7
Online ISBN: 978-3-319-01168-4
eBook Packages: EngineeringEngineering (R0)