RESUMEN
Fibers of liquid crystal elastomers (LCEs) as promising artificial muscle show ultralarge and reversible contractile strokes. However, the contractile force is limited by the poor mechanical properties of the LCE fibers. Herein, we report high-strength LCE fibers by introducing a secondary network into the single-network LCE. The double-network LCE (DNLCE) shows considerable improvements in tensile strength (313.9%) and maximum actuation stress (342.8%) compared to pristine LCE. To facilitate the controllability and application, a coiled artificial muscle fiber consisting of DNLCE-coated carbon nanotube (CNT) fiber is prepared. When electrothermally driven, the artificial muscle fiber outputs a high actuation performance and programmable actuation. Furthermore, by knitting the artificial muscle fibers into origami structures, an intelligent gripper and crawling inchworm robot have been demonstrated. These demonstrations provide promising application scenarios for advanced intelligent systems in the future.
RESUMEN
Generative Flow Networks (GFlowNets) aim to generate diverse trajectories from a distribution in which the final states of the trajectories are proportional to the reward, serving as a powerful alternative to reinforcement learning for exploratory control tasks. However, the individual-flow matching constraint in GFlowNets limits their applications for multi-agent systems, especially continuous joint-control problems. In this paper, we propose a novel Multi-Agent generative Continuous Flow Networks (MACFN) method to enable multiple agents to perform cooperative exploration for various compositional continuous objects. Technically, MACFN trains decentralized individual-flow-based policies in a centralized global-flow-based matching fashion. During centralized training, MACFN introduces a continuous flow decomposition network to deduce the flow contributions of each agent in the presence of only global rewards. Then agents can deliver actions solely based on their assigned local flow in a decentralized way, forming a joint policy distribution proportional to the rewards. To guarantee the expressiveness of continuous flow decomposition, we theoretically derive a consistency condition on the decomposition network. Experimental results demonstrate that the proposed method yields results superior to the state-of-the-art counterparts and better exploration capability. Our code is available at https://github.com/isluoshuang/MACFN.
Asunto(s)
Aprendizaje , Políticas , Refuerzo en Psicología , RecompensaRESUMEN
The development of wearable electronics has driven the need for smart fibers with advanced multifunctional synergy. In this paper, we present a design of a multifunctional coaxial fiber that is composed of a biopolymer-derived core and an MXene/silver nanowire (AgNW) sheath by wet spinning. The fiber synergistically integrates moisture actuation, length tracing, humidity sensing, and electric heating, making it highly promising for portable devices and protective systems. The biopolymer-derived core provides deformation for moisture-sensitive actuation, while the MXene/AgNW sheath with good conductivity enables the fiber to perform electric heating, humidity sensing, and self-sensing actuation. The coaxial fiber can be programmed to rapidly desorb water molecules to shrink to its original length by using the MXene/AgNW sheath as an electrical heater. We demonstrate proof-of-concept applications based on the multifunctional fibers for thermal physiotherapy and wound healing/monitoring. The sodium alginate@MXene-based coaxial fiber presents a promising solution for the next-generation of smart wearable electronics.