Skilled robotic manipulation benefits from complex synergies between non-prehensile (e.g. pushing) and prehensile grasping) actions: pushing can help rearrange cluttered objects to make space for arms fingers; likewise, grasping displace movements more precise collision-free. In this work, we demonstrate that it is possible discover learn these scratch through model-free deep reinforcement learning. Our method involves training two fully convolutional networks map visual observations one...

We investigate whether a robot arm can learn to pick and throw arbitrary rigid objects into selected boxes quickly accurately. Throwing has the potential increase physical reachability picking speed of arm. However, precisely throwing in unstructured settings presents many challenges: from acquiring grasps suitable for reliable throwing, handling varying object-centric properties (e.g., mass distribution, friction, shape) complex aerodynamics. In this work, we propose an end-to-end...

Transparent objects are a common part of everyday life, yet they possess unique visual properties that make them incredibly difficult for standard 3D sensors to produce accurate depth estimates for. In many cases, often appear as noisy or distorted approximations the surfaces lie behind them. To address these challenges, we present ClearGrasp - deep learning approach estimating geometry transparent from single RGB-D image robotic manipulation. Given objects, uses convolutional networks infer...

Large pretrained (e.g., "foundation") models exhibit distinct capabilities depending on the domain of data they are trained on. While these domains generic, may only barely overlap. For example, visual-language (VLMs) Internet-scale image captions, but large language (LMs) further text with no images spreadsheets, SAT questions, code). As a result, store different forms commonsense knowledge across domains. In this work, we show that diversity is symbiotic, and can be leveraged through...

We investigate whether a robot arm can learn to pick and throw arbitrary objects into selected boxes quickly accurately.Throwing has the potential increase physical reachability picking speed of arm.However, precisely throwing in unstructured settings presents many challenges: from acquiring reliable pre-throw conditions (e.g.grasp object) handling varying object-centric properties (e.g.mass distribution, friction, shape) dynamics (e.g.aerodynamics).In this work, we propose an end-to-end...

Is it possible to learn policies for robotic assembly that can generalize new objects? We explore this idea in the context of kit task. Since classic methods rely heavily on object pose estimation, they often struggle objects without 3D CAD models or task-specific training data. In work, we propose formulate task as a shape matching problem, where goal is descriptor establishes geometric correspondences between surfaces and their target placement locations from visual input. This formulation...

This paper describes how to instrument the physical world so that objects become self-describing, communicating their identity, geometry, and other information such as history or user annotation. The enabling technology is a wireless tag which acts radio frequency identity geometry (RFIG) transponder. We show addition of photo-sensor significantly extends its functionality allow geometric operations - finding 3D position tag, detecting change in shape tagged object. Tag data presented by...

Introduction Is stickiness the Holy Grail for e-tailing? In general, refers to amount of time a person spends on Web site during visiting session (such as, ) or over specified period ). Zauberman equates and "within-site lock-in" uses it approximate visitors' loyalty site. The conventional wisdom suggests that is crucial can contribute e-tailers' bottom lines considerably. However, direct economic impacts have not been duly examined empirically, particularly from perspective consumers'...

We find that across a wide range of robot policy learning scenarios, treating supervised with an implicit model generally performs better, on average, than commonly used explicit models. present extensive experiments this finding, and we provide both intuitive insight theoretical arguments distinguishing the properties models compared to their counterparts, particularly respect approximating complex, potentially discontinuous multi-valued (set-valued) functions. On robotic tasks show...

Typical end-to-end formulations for learning robotic navigation involve predicting a small set of steering command actions (e.g., step forward, turn left, right, etc.) from images the current state bird's-eye view SLAM reconstruction). Instead, we show that it can be advantageous to learn with dense action representations defined in same domain as state. In this work, present "spatial maps," which possible is represented by pixel map (aligned input image state), where each represents local...

This paper deals with the vibration-to-electrical transducer that has an M-size form factor and generates a DC voltage can power off-the-shelf integrated circuits. Vibration-powered wireless sensors obtain from machine vibrations, human movement, or other forms of motion. The feasibility incorporating micro transducers (MPTs) multiplier rectifier to make generator (MPG) is demonstrated, same size shape as AA battery. AA-sized module includes large capacitor produce output

Kinetic (dynamic) typography has demonstrated the ability to add significant emotive content and appeal expressive text, allowing some of qualities normally found in film spoken word be added static text. been widely successfully used title sequences as well television computer-based advertising. However, its communicative abilities have not studied, potential rarely exploited outside these areas. This is partly due difficulty creating kinetic with current tools, often requiring hours work...

A number of projects within the computer graphics, vision, and human-computer interaction communities have recognized value using projected structured light patterns for purposes doing range finding, location dependent data delivery, projector adaptation, or object discovery tracking. However, most work exploring these concepts has relied on visible resulting in a caustic visual experience. In this work, we present first design implementation high-resolution, scalable, general purpose...

Computational fluid dynamics (CFD) is a branch of mechanics that uses numerical analysis and data structures to calculate, analyze, visualize (liquids, gases, dissolved gases) flows. This document provides general introductions best practices for CFD modeling in water infrastructure practitioners, particularly those new modeling, which becoming widely used tool the design retrofitting water, wastewater, stormwater infrastructure. The method serves as an alternative, or complement, physical...

The objective of this paper is to present an experimentally validated mechanistic model predict the oxygen transfer rate coefficient (Kla) in aeration tanks for different water temperatures. Using experimental data created by Hunter and Vogelaar, formula precisely reproduces results standardized Kla at 20 °C, comparatively better than current used ASCE 2-06 based on equation Kla20 = Kla. ([Formula: see text])(20-T) where T °C. Currently, reported values [Formula: text] range from 1.008...

Autonomous fabric manipulation is a longstanding challenge in robotics, but evaluating progress difficult due to the cost and diversity of robot hardware. Using Reach, cloud robotics platform that enables low-latency remote execution control policies on physical robots, we present first systematic benchmarking al-gorithms We develop 4 novel learning-based algorithms model expert actions, keypoints, reward functions, dynamic motions, compare these against learning-free inverse dynamics task...

We investigate whether a robot arm can learn to pick and throw arbitrary objects into selected boxes quickly accurately. Throwing has the potential increase physical reachability picking speed of arm. However, precisely throwing in unstructured settings presents many challenges: from acquiring reliable pre-throw conditions (e.g. initial pose object manipulator) handling varying object-centric properties mass distribution, friction, shape) dynamics aerodynamics). In this work, we propose an...

Over 60 percent of all wastewater treatment plants in the developed countries use activated sludge process as their secondary system. About 50 to 85 total energy consumed a biological plant is aeration. The process, most common such performed large aeration basins provide air for microorganisms remove nutrients and pollutants through biodegradation. Designs supply often fail meet sustained peak organic loading, which may lead unsatisfactory performance even failure. On other hand, excessive...

The ideas of ubiquitous always‐on AR glasses that enhance your visual experience the world is an exciting one. While industry continues to mature hardware and software technologies necessary for compelling consumer HMDs, we will share what possible today in a form factor you carry with everyday: mobile phone. We touch on algorithms sensors enable end‐user facing applications such as measuring tools, shopping, games, indoor positioning.

Transparent objects are a common part of everyday life, yet they possess unique visual properties that make them incredibly difficult for standard 3D sensors to produce accurate depth estimates for. In many cases, often appear as noisy or distorted approximations the surfaces lie behind them. To address these challenges, we present ClearGrasp -- deep learning approach estimating geometry transparent from single RGB-D image robotic manipulation. Given objects, uses convolutional networks...

Is it possible to learn policies for robotic assembly that can generalize new objects? We explore this idea in the context of kit task. Since classic methods rely heavily on object pose estimation, they often struggle objects without 3D CAD models or task-specific training data. In work, we propose formulate task as a shape matching problem, where goal is descriptor establishes geometric correspondences between surfaces and their target placement locations from visual input. This formulation...