Details
| Original language | English |
|---|---|
| Pages (from-to) | 1400-1407 |
| Number of pages | 8 |
| Journal | Robotics and autonomous systems |
| Volume | 60 |
| Issue number | 11 |
| Publication status | Published - Nov 2012 |
| Externally published | Yes |
Abstract
Reinforcement learning (RL) is a biologically supported learning paradigm, which allows an agent to learn through experience acquired by interaction with its environment. Its potential to learn complex action sequences has been proven for a variety of problems, such as navigation tasks. However, the interactive randomized exploration of the state space, common in reinforcement learning, makes it difficult to be used in real-world scenarios. In this work we describe a novel real-world reinforcement learning method. It uses a supervised reinforcement learning approach combined with Gaussian distributed state activation. We successfully tested this method in two real scenarios of humanoid robot navigation: first, backward movements for docking at a charging station and second, forward movements to prepare grasping. Our approach reduces the required learning steps by more than an order of magnitude, and it is robust and easy to be integrated into conventional RL techniques.
Keywords
- Autonomous docking, Humanoid robots, Real-world, Reinforcement learning, SARSA
ASJC Scopus subject areas
- Engineering(all)
- Control and Systems Engineering
- Computer Science(all)
- Software
- Mathematics(all)
- General Mathematics
- Computer Science(all)
- Computer Science Applications
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
In: Robotics and autonomous systems, Vol. 60, No. 11, 11.2012, p. 1400-1407.
Research output: Contribution to journal › Article › Research › peer review
}
TY - JOUR
T1 - Real-world reinforcement learning for autonomous humanoid robot docking
AU - Navarro-Guerrero, Nicolás
AU - Weber, Cornelius
AU - Schroeter, Pascal
AU - Wermter, Stefan
PY - 2012/11
Y1 - 2012/11
N2 - Reinforcement learning (RL) is a biologically supported learning paradigm, which allows an agent to learn through experience acquired by interaction with its environment. Its potential to learn complex action sequences has been proven for a variety of problems, such as navigation tasks. However, the interactive randomized exploration of the state space, common in reinforcement learning, makes it difficult to be used in real-world scenarios. In this work we describe a novel real-world reinforcement learning method. It uses a supervised reinforcement learning approach combined with Gaussian distributed state activation. We successfully tested this method in two real scenarios of humanoid robot navigation: first, backward movements for docking at a charging station and second, forward movements to prepare grasping. Our approach reduces the required learning steps by more than an order of magnitude, and it is robust and easy to be integrated into conventional RL techniques.
AB - Reinforcement learning (RL) is a biologically supported learning paradigm, which allows an agent to learn through experience acquired by interaction with its environment. Its potential to learn complex action sequences has been proven for a variety of problems, such as navigation tasks. However, the interactive randomized exploration of the state space, common in reinforcement learning, makes it difficult to be used in real-world scenarios. In this work we describe a novel real-world reinforcement learning method. It uses a supervised reinforcement learning approach combined with Gaussian distributed state activation. We successfully tested this method in two real scenarios of humanoid robot navigation: first, backward movements for docking at a charging station and second, forward movements to prepare grasping. Our approach reduces the required learning steps by more than an order of magnitude, and it is robust and easy to be integrated into conventional RL techniques.
KW - Autonomous docking
KW - Humanoid robots
KW - Real-world
KW - Reinforcement learning
KW - SARSA
UR - http://www.scopus.com/inward/record.url?scp=84867874785&partnerID=8YFLogxK
U2 - 10.1016/j.robot.2012.05.019
DO - 10.1016/j.robot.2012.05.019
M3 - Article
AN - SCOPUS:84867874785
VL - 60
SP - 1400
EP - 1407
JO - Robotics and autonomous systems
JF - Robotics and autonomous systems
SN - 0921-8890
IS - 11
ER -