The integration of multi-agent reinforcement learning (MARL) into complex systems has paved new ways for collaborative problem-solving. However, traditional approaches to MARL frequently encounter the challenge of ach...
详细信息
ISBN:
(数字)9798350368604
ISBN:
(纸本)9798350368611
The integration of multi-agent reinforcement learning (MARL) into complex systems has paved new ways for collaborative problem-solving. However, traditional approaches to MARL frequently encounter the challenge of achieving ef-ficient communication among agents, essential for coordinated action. This paper introduces a region division and leader-follower(RDLF) communication algrithm with the MARL frame-work. RDLF divides the environment into several regions, each managed by a leader agent that coordinates the actions of fol-lower agents and handling inter-region communication. This hier-archical structure reduces unnecessary communication, enhancing learning efficiency. Experimental results in multi-particle en-vironments demonstrate RDLF's superiority over existing MARL algorithms, especially with increasing agent numbers. RDLF effectively addresses scalability and communication challenges in large-scale multi-agent systems, providing a robust foundation for its application in complex and dynamic environment.
This paper presents a novel learning-based method for a robotic manipulator to achieve liquid pouring across different liquid levels using only visual sensors. Previous works have relied on either online reinforcement...
详细信息
ISBN:
(数字)9798331509644
ISBN:
(纸本)9798331509651
This paper presents a novel learning-based method for a robotic manipulator to achieve liquid pouring across different liquid levels using only visual sensors. Previous works have relied on either online reinforcement learning or imitation learning, which are limited by sim-to-real gaps or data efficiency. In this paper, we propose to combine supervised learning and offline reinforcement learning, utilizing a human demonstration dataset containing only successful pouring at the highest level. Specifically, our approach employs supervised learning for a visual classifier, transforming the visual input into a categorical distribution over liquid levels. The offline reinforcement learning method is applied to a binary-conditioned control policy, which takes a binary signal and the robot's proprioception as inputs to determine the action. The binary signal indicates whether the desired liquid level has been reached based on the target liquid level. Through experiments, our method demonstrates superior effectiveness compared to a state-of-the-art imitation learning algorithm. Moreover, tests in unseen scenarios and multiple liquid level commands verify the generalization and transferability of our approach.
In this paper, we present a Q-learning algorithm for finite-horizon H8 tracking control of unknown linear discrete-time systems. Finite horizon control is challenging due to its correspondence with a time-varying Ricc...
详细信息
The effect of the Si addition content ranging from 3.3 to 9.1 at.% on the microstructure, mechanical properties and LBE corrosion behaviour of AlCrFeMoTi HEA coatings was investigated. The AlCrFeMoTiSix coatings still...
详细信息
The effect of the Si addition content ranging from 3.3 to 9.1 at.% on the microstructure, mechanical properties and LBE corrosion behaviour of AlCrFeMoTi HEA coatings was investigated. The AlCrFeMoTiSix coatings still...
详细信息
Two kinds of semi-solid samples of AZ80−0.2Y−0.15Ca(wt.%)(AZ80M)magnesium alloy were prepared by semi-solid isothermal heat treatment of materials with and without equal channel angular pressing(ECAP)*** microstructur...
详细信息
Two kinds of semi-solid samples of AZ80−0.2Y−0.15Ca(wt.%)(AZ80M)magnesium alloy were prepared by semi-solid isothermal heat treatment of materials with and without equal channel angular pressing(ECAP)*** microstructures of initial and semi-solid treated samples were compared and *** results showed a significant difference in the liquid phase distribution between three-pass ECAP processed(3P)and as-received samples during the isothermal heating *** semi-solid 3P sample showed a more uniform liquid distribution due to its smaller dihedral ***,the coarsening processes of solid grains of as-received and 3P samples were dominated by the coalescence and Ostwald ripening mechanism,*** difference of coarsening processes was mainly related to the proportion of the high-angle grain boundaries in materials,which further affected the evolution behavior of the liquid pools.
Density wave oscillation (DWO) in parallel channels is an important problem that has been widely studied. The marginal stability boundary (MSB) is often found to be "L " shaped on the N pch -N sub plane, but...
详细信息
The (HEA/HEAN)-n multilayer coatings consisting of alternating (AlCrFeMoTi)N (top) and AlCrFeMoTi layers were specially designed and successfully deposited on F/M steel substrate by reactive magnetron sputtering techn...
详细信息
Experimental study of coherent structures downstream of a spacer grid with different inclination angles of mixing vanes in a 5×5 rod bundle was conducted by TR-PIV under four Reynolds *** coherent structures down...
详细信息
The effect of the Si addition content ranging from 3.3 to 9.1 at.% on the microstructure, mechanical properties and LBE corrosion behaviour of AlCrFeMoTi HEA coatings was investigated. The AlCrFeMoTiSix coatings still...
详细信息
暂无评论