Search for: All records

3D perceptual representations are well suited for robot manipulation as they easily encode occlusions and simplify spatial reasoning. Many manipulation tasks require high spatial precision in end-effector pose prediction, which typically demands high-resolution 3D feature grids that are computationally expensive to process. As a result, most manipulation policies operate directly in 2D, foregoing 3D inductive biases. In this paper, we introduce Act3D, a manipulation policy transformer that represents the robot’s workspace using a 3D feature field with adaptive resolutions dependent on the task at hand. The model lifts 2D pre-trained features to 3D using sensed depth, and attends to them to compute features for sampled 3D points. It samples 3D point grids in a coarse to fine manner, featurizes them using relative-position attention, and selects where to focus the next round of point sampling. In this way, it efficiently computes 3D action maps of high spatial resolution. Act3D sets a new state-of-the-art in RLBench, an established manipulation benchmark, where it achieves 10% absolute improvement over the previous SOTA 2D multi-view policy on 74 RLBench tasks and 22% absolute improvement with 3x less compute over the previous SOTA 3D policy. We quantify the importance of relative spatial attention, large-scale vision-language pre-trained 2D backbones, and weight tying across coarse-to-fine attentions in ablative experiments. Code and videos are available at our project site: https://act3d.github.io/. 

Rechargeable Li-CO2 batteries have emerged as promising candidates for next generation batteries due to their low cost, high theoretical capacity, and ability to capture the greenhouse gas CO2. However, these batteries still face challenges such as slow reaction kinetic and short cycle performance due to the accumulation of discharge products. To address this issue, it is necessary to design and develop high efficiency electrocatalysts that can improve CO2 reduction reaction. In this study, we report the use of NiMn2O4 electrocatalysts combined with multiwall carbon nanotubes as a cathode material in the Li-CO2 batteries. This combination proved effective in decomposing discharge products and enhancing cycle performance. The battery shows stable discharge–charge cycles for at least 30 cycles with a high limited capacity of 1000 mAh/g at current density of 100 mA/g. Furthermore, the battery with the NiMn2O4@CNT catalyst exhibits a reversible discharge capacity of 2636 mAh/g. To gain a better understanding of the reaction mechanism of Li-CO2 batteries, spectroscopies and microscopies were employed to identify the chemical composition of the discharge products. This work paves a pathway to increase cycle performance in metal-CO2 batteries, which could have significant implications for energy storage and the reduction of greenhouse gas emissions. 

The next generation of fuel cells, electrolyzers, and batteries requires higher power, faster kinetics, and larger energy density, which necessitate the use of compositionally complex oxides to achieve multifunctionalities and activity. These compositionally complex oxides may change their phases and structures during an electrochemical process—a so-called “electrochemically driven phase transformation.” The origin for such a phase change has remained obscure. The aim of this paper is to present an experimental study and a theoretical analysis of phase evolution in praseodymium nickelates. Nickelate-based electrodes show up to 60 times greater phase transformation during operation when compared with thermally annealed ones. Theoretical analysis suggests that the presence of a reduced oxygen partial pressure at the interface between the oxygen electrode and the electrolyte is the origin for the phase change in an oxygen electrode. Guided by the theory, the addition of the electronic conduction in the interface layer leads to the significant suppression of phase change while improving cell performance and performance stability. 

The neglected tropical disease schistosomiasis impacts over 700 million people globally. Schistosoma mansoni , the trematode parasite that causes the most common type of schistosomiasis, requires planorbid pond snails of the genus Biomphalaria to support its larval development and transformation to the cercarial form that can infect humans. A greater understanding of neural signaling systems that are specific to the Biomphalaria intermediate host could lead to novel strategies for parasite or snail control. This study examined a Biomphalaria glabrata neural channel that is gated by the neuropeptide FMRF-NH 2 . The Biomphalaria glabrata FMRF-NH 2 gated sodium channel ( Bgl- FaNaC) amino acid sequence was highly conserved with FaNaCs found in related gastropods, especially the planorbid Planorbella trivolvis (91% sequence identity). In common with the P . trivolvis FaNaC, the B . glabrata channel exhibited a low affinity (EC 50 : 3 x 10 −4 M) and high specificity for the FMRF-NH 2 agonist. Its expression in the central nervous system, detected with immunohistochemistry and in situ hybridization, was widespread, with the protein localized mainly to neuronal fibers and the mRNA confined to cell bodies. Colocalization of the Bgl- FaNaC message with its FMRF-NH 2 agonist precursor occurred in some neurons associated with male mating behavior. At the mRNA level, Bgl- FaNaC expression was decreased at 20 and 35 days post infection (dpi) by S . mansoni . Increased expression of the transcript encoding the FMRF-NH 2 agonist at 35 dpi was proposed to reflect a compensatory response to decreased receptor levels. Altered FMRF-NH 2 signaling could be vital for parasite proliferation in its intermediate host and may therefore present innovative opportunities for snail control.