Accelerating decentralized reinforcement learning of complex individual behaviors

Leottau, David Leonardo; Lobos-Tsunekawa, Kenzo; Jaramillo, Francisco; Ruiz‐del‐Solar, Javier

doi:10.1016/j.engappai.2019.06.019

Cited by 5 publications

(3 citation statements)

References 5 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Reinforcement learning for robot motion control suffers from the complexity of action space increases exponentially with the number of robot actuators, which makes policy search and optimization difficult (Leottau et al (2018); Leottau et al (2019)). Decentralized reinforcement learning methods decompose the robot into multiple agents to explore the action space coordinately to achieve the common goal.…”

Section: Decentralized Reinforcement Learningmentioning

confidence: 99%

“…Decentralized reinforcement learning methods decompose the robot into multiple agents to explore the action space coordinately to achieve the common goal. The decomposed action space will improve policy learning efficiency by reducing the search space of policies (Leottau et al (2019)). However, the decentralized learning system has non-stationary and non-Markovian issues, which causes policy divergence, especially in the multitask situation.…”

Section: Decentralized Reinforcement Learningmentioning

confidence: 99%

See 1 more Smart Citation

Decentralized Policy Learning for Quadruped Robots Multi-Gait Control

Zhang

2023

Preprint

View full text Add to dashboard Cite

This paper proposes the Decentralized Multi-Critic Actor Learning (DMultiCriticAL) method for quadruped robots to overcome the multi-task gradient conflict in multi-gait learning. In this method, each leg joint angle of quadruped robots is served as the action space of decentralized learning agents rather than learning directly to control all robot joint angles. Because of the high similarity in quadruped robot local action space, the DMultiCriticAL method can avoid the multi-task gradient conflict in multi-gait learning. Compared with the conventional multi-task learning method, we exploit the decentralized learning systems to limit the useless exploration of environments by agent, not design complex reward functions to track the multi-gait reference motion trajectory. This also means we can avoid the suboptimal reference motion trajectory design. Our experiment results show that the multi-gait motion controller based on the DMultiCriticAL method can yield more stable locomotion without complex reward function designs.

show abstract

Section: Decentralized Reinforcement Learningmentioning

confidence: 99%

Section: Decentralized Reinforcement Learningmentioning

confidence: 99%

Decentralized Policy Learning for Quadruped Robots Multi-Gait Control

Zhang

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…In recent decades, and with the advancement in artificial intelligence (AI) models, they became powerful tools to solve complex engineering problems [19][20][21][22][23][24][25][26][27][28][29] . A number of AI-based studies have conducted to enhance the accuracy of Dx estimation in turbulent flow systems such as streams 30,33 .…”

Section: Introductionmentioning

confidence: 99%

Uncertainty Quantification of Granular Computing-neural Network Model for Prediction of Pollutant Longitudinal Dispersion Coefficient in Aquatic Streams

Ghiasi

Yuanbin

Noori

et al. 2021

Preprint

View full text Add to dashboard Cite

Discharge of pollution loads into natural water systems remains a global challenge that threatens water/food supply as well as endangers ecosystem services. Natural rehabilitation of the polluted streams is mainly influenced by the rate of longitudinal dispersion (Dx), a key parameter with large temporal and spatial fluctuates that characterizes pollution transport. The large uncertainty in estimation of Dx in streams limits evaluation of water quality in natural streams and design of water quality enhancement strategies. This study develops a sophisticated model coupled with granular computing and neural network models (GrC-ANN) to provide robust prediction of Dx and its uncertainty for different flow-geometric conditions with high spatiotemporal variability. Uncertainty analysis of Dx GrC-ANN model was based on the alteration of training data fed to tune the model. Modified bootstrap method was employed to generate different training patterns through resampling from a 503 global database of tracer experiments in streams. Comparison between the Dx values estimated by GrC-ANN to those determined from tracer measurements show the appropriateness and robustness of the proposed method in determining the rate of longitudinal dispersion. GrC-ANN model with the narrowest bandwidth of estimated uncertainty (bandwidth-factor =0.56) that brackets the most percentage of true Dx data (i.e., 100%) is the best model to compute Dx in streams. Given considerable inherent uncertainty reported in other Dx models, the Dx GrC-ANN model is suggested as a proper tool for further studies of pollutant mixing in turbulent flow systems such as streams.

show abstract

The Barzilai–Borwein Method for distributed optimization over unbalanced directed networks

Chen

Zheng

et al. 2021

Engineering Applications of Artificial Intelligence

View full text Add to dashboard Cite

Accelerating decentralized reinforcement learning of complex individual behaviors

Cited by 5 publications

References 5 publications

Decentralized Policy Learning for Quadruped Robots Multi-Gait Control

Decentralized Policy Learning for Quadruped Robots Multi-Gait Control

Uncertainty Quantification of Granular Computing-neural Network Model for Prediction of Pollutant Longitudinal Dispersion Coefficient in Aquatic Streams

The Barzilai–Borwein Method for distributed optimization over unbalanced directed networks

Contact Info

Product

Resources

About