Publication detail

Monte Carlo tree search control scheme for multibody dynamics applications

Yixuan Tang Grzegorz Orzechowski Aleš Prokop Aki Mikkola

Czech title

Schéma řízení pro multibody modely na základě Monte Carlo Tree Search

English title

Monte Carlo tree search control scheme for multibody dynamics applications

Type

journal article in Web of Science

Language

en

Original abstract

There is considerable interest in applying reinforcement learning (RL) to improve machine control across multiple industries, and the automotive industry is one of the prime examples. Monte Carlo Tree Search (MCTS) has emerged and proven powerful in decision-making games, even without understanding the rules. In this study, multibody system dynamics (MSD) control is first modeled as a Markov Decision Process and solved with Monte Carlo Tree Search. Based on randomized search space exploration, the MCTS framework builds a selective search tree by repeatedly applying a Monte Carlo rollout at each child node. However, without a library of available choices, deciding among the many possibilities for agent parameters can be intimidating. In addition, the MCTS poses a significant challenge for searching due to the large branching factor. This challenge is typically overcome by appropriate parameter design, search guiding, action reduction, parallelization, and early termination. To address these shortcomings, the overarching goal of this study is to provide needed insight into inverted pendulum controls via vanilla and modified MCTS agents, respectively. A series of reward functions are well-designed according to the control goal, which maps a specific distribution shape of reward bonus and guides the MCTS-based control to maintain the upright position. Numerical examples show that the reward-modified MCTS algorithms significantly improve the control performance and robustness of the default choice of a constant reward that constitutes the vanilla MCTS. The exponentially decaying reward functions perform better than the constant value or polynomial reward functions. Moreover, the exploitation vs. exploration trade-off and discount parameters are carefully tested. The study’s results can guide the research of RL-based MSD users.

English abstract

There is considerable interest in applying reinforcement learning (RL) to improve machine control across multiple industries, and the automotive industry is one of the prime examples. Monte Carlo Tree Search (MCTS) has emerged and proven powerful in decision-making games, even without understanding the rules. In this study, multibody system dynamics (MSD) control is first modeled as a Markov Decision Process and solved with Monte Carlo Tree Search. Based on randomized search space exploration, the MCTS framework builds a selective search tree by repeatedly applying a Monte Carlo rollout at each child node. However, without a library of available choices, deciding among the many possibilities for agent parameters can be intimidating. In addition, the MCTS poses a significant challenge for searching due to the large branching factor. This challenge is typically overcome by appropriate parameter design, search guiding, action reduction, parallelization, and early termination. To address these shortcomings, the overarching goal of this study is to provide needed insight into inverted pendulum controls via vanilla and modified MCTS agents, respectively. A series of reward functions are well-designed according to the control goal, which maps a specific distribution shape of reward bonus and guides the MCTS-based control to maintain the upright position. Numerical examples show that the reward-modified MCTS algorithms significantly improve the control performance and robustness of the default choice of a constant reward that constitutes the vanilla MCTS. The exponentially decaying reward functions perform better than the constant value or polynomial reward functions. Moreover, the exploitation vs. exploration trade-off and discount parameters are carefully tested. The study’s results can guide the research of RL-based MSD users.

Keywords in English

Monte Carlo Tree Search; Multibody dynamics; Reward functions; Parametric analysis; Artificial intelligence control; Inverted pendulum

Released

03.04.2024

Publisher

Springer Nature

ISSN

1573-269X

Volume

112

Number

10

Pages from–to

8363–8391

Pages count

28

BIBTEX


@article{BUT188384,
  author="Yixuan {Tang} and Grzegorz {Orzechowski} and Aleš {Prokop} and Aki {Mikkola},
  title="Monte Carlo tree search control scheme for multibody dynamics applications",
  year="2024",
  volume="112",
  number="10",
  month="April",
  pages="8363--8391",
  publisher="Springer Nature",
  issn="1573-269X"
}