Consideration on robotic giant-swing motion generated by reinforcement learning

Hara, Masayuki; Kawabe, Naoto; Sakai, Naoki; Huang, Jian; Bleuler, Hannes; Yabuta, Tetsuro

doi:10.1109/iros.2009.5354270

Cited by 13 publications

(6 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The earlier phase is the initial phase around the position right below the bar, and in the later phase the robot reaches the angle that is large enough to achieve giant-swing. Hara et al [8] and Sakai et al [9] have confirmed the need to distinguish the two phases and they have realized the substantial distinction by giving the reward of two levels and switching between greedy and random actions. However, it is better if we do not need to make such a distinction and the robot more actively learn how to act.…”

Section: Discussionmentioning

confidence: 94%

“…3) by reinforcement learning with decision-making using LS. Many researchers have studied this kind of robot, which imitates high bar gymnastics, as a control test bed with nonlinear dynamics [7][8][9][17][18][19]. As in Fig.…”

Section: Giant-swing Robotmentioning

confidence: 99%

“…In this study, we construct a giant-swing simulator with Open Dynamics Engine (ODE). Consulting the works by Hara et al [8,9], the two links are both set to have 0.2 m and 0.5 kg. The active hip joint moves at angular velocity 4.0 rad/s in the operational range of ሾͲǡ ሺͷ Τ ሻ ߨሿ rad.…”

Section: Giant-swing Robotmentioning

confidence: 99%

“…Hara et al [8], Sakai et al [9] and Toyoda et al [10] have studied the problems of simply applying Q-learning method, that is the representative algorithm for reinforcement learning, to the real robots. They coarse-grained the state space in the giant-swing motion learning so as to make Q-learning practicable.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

The efficacy of symmetric cognitive biases in robotic motion learning

Uragami

Takahashi

Alsubeheen

et al. 2011

2011 IEEE International Conference on Mechatronics and Automation

View full text Add to dashboard Cite

We propose an application of human-like decision-making to robotic motion learning. Human is known to have illogical symmetric cognitive biases that induce "if p then q" and "if not q then not p" from "if q then p." The loosely symmetric Shinohara model quantitatively represents the tendencies (Shinohara et al. 2007). Previous studies one of the authors have revealed that an agent with the model used as the action value function shows great performance in n-armed bandit problems, because of the illogical biases. In this study, we apply the model to reinforcement learning with Q-learning algorithm. Testing the model on a simulated giant-swing robot, we have confirmed its efficacy in convergence speed increase and avoidance of local optimum.

show abstract

Section: Discussionmentioning

confidence: 94%

Section: Giant-swing Robotmentioning

confidence: 99%

Section: Giant-swing Robotmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

The efficacy of symmetric cognitive biases in robotic motion learning

Uragami

Takahashi

Alsubeheen

et al. 2011

2011 IEEE International Conference on Mechatronics and Automation

View full text Add to dashboard Cite

show abstract

“…Ono and Yamaura (5) - (7) proposed a feedback control realized by configuration control to follow free giant swing motions which were derived by the optimum trajectory planning method. Hara (8) attempted to make a compact humanoid robot acquire a giant-swing motion without any robotic models by using reinforcement learning. However, there has been shown its complexity when seeking a feedback gain or calculating an accurate target trajectory.…”

Section: Introductionmentioning

confidence: 99%

Giant Swing Motion Control of 3-link Gymnastic Robot with Friction around an Underactuated Joint

Liu

Yamaura

2011

JSDD

View full text Add to dashboard Cite

This paper discusses a control method which can stabilize the giant swing motions of 3-link horizontal bar gymnastic robot. A method called Multiple-prediction Delayed Feedback Control(MDFC) proposed by the authors, which has been shown to be effective in control such chaos system, is extended in this paper to experimental gymnastic robot system, taking in account the effect of friction around the link joint. The dynamic of underactuated systems like gymnastic robot shows significant differences in cases of with friction and without friction. However, MDFC considers only the ideal situation without friction. Therefore, a stabilization control method consisting of three kind of control inputs which the one is MDFC for guaranteeing asymptotic stability and the others are used for friction compensation is derived. Numerical simulations and experimental results prove its effectiveness of our proposed method.

show abstract