Search for: All records

Network Markov Decision Processes (MDPs), which are the de-facto model for multi-agent control, pose a significant challenge to efficient learning caused by the exponential growth of the global state-action space with the number of agents. In this work, utilizing the exponential decay property of network dynamics, we first derive scalable spectral local representations for multiagent reinforcement learning in network MDPs, which induces a network linear subspace for the local $$Q$$-function of each agent. Building on these local spectral representations, we design a scalable algorithmic framework for multiagent reinforcement learning in continuous state-action network MDPs, and provide end-to-end guarantees for the convergence of our algorithm. Empirically, we validate the effectiveness of our scalable representation-based approach on two benchmark problems, and demonstrate the advantages of our approach over generic function approximation approaches to representing the local $$Q$$-functions. 

Wire arc additive manufacturing (WAAM) has received increasing use in 3D printing because of its high deposition rates suitable for components with large and complex geometries. However, the lower forming accuracy of WAAM than other metal additive manufacturing methods has imposed limitations on manufacturing components with high precision. To resolve this issue, we herein implemented the hybrid manufacturing (HM) technique, which integrated WAAM and subtractive manufacturing (via a milling process), to attain high forming accuracy while taking advantage of both WAAM and the milling process. We describe in this paper the design of a robot-based HM platform in which the WAAM and CNC milling are integrated using two robotic arms: one for WAAM and the other for milling immediately following WAAM. The HM was demonstrated with a thin-walled aluminum 5356 component, which was inspected by X-ray micro-computed tomography (μCT) for porosity visualization. The temperature and cutting forces in the component under milling were acquired for analysis. The surface roughness of the aluminum component was measured to assess the surface quality. In addition, tensile specimens were cut from the components using wire electrical discharge machining (WEDM) for mechanical testing. Both machining quality and mechanical properties were found satisfactory; thus the robot-based HM platform was shown to be suitable for manufacturing high-quality aluminum parts. 

Abstract We introduce a new method based on nonnegative matrix factorization, Neural NMF, for detecting latent hierarchical structure in data. Datasets with hierarchical structure arise in a wide variety of fields, such as document classification, image processing, and bioinformatics. Neural NMF recursively applies NMF in layers to discover overarching topics encompassing the lower-level features. We derive a backpropagation optimization scheme that allows us to frame hierarchical NMF as a neural network. We test Neural NMF on a synthetic hierarchical dataset, the 20 Newsgroups dataset, and the MyLymeData symptoms dataset. Numerical results demonstrate that Neural NMF outperforms other hierarchical NMF methods on these data sets and offers better learned hierarchical structure and interpretability of topics.