Abstract: Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for visually detecting a halocline. In some implementations, a method includes moving a camera through different depths of water within a fish enclosure, capturing, at the different depths, images of fish, determining that changes in focus in the images correspond to changes in depth that the images were captured, and based on determining that the changes in focus in the images correspond to the changes in depths that the images were captured, detecting a halocline at a particular depth.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for a device with improved ambient temperature detection. In some implementations, a device includes a housing that forms an interior space, and the housing includes an exterior surface, a pass-through region that defines a through-hole between the interior space to the exterior surface, and a recess at the exterior surface adjacent to the pass-through hole. The device includes a printed circuit board disposed within the interior space of the plastic housing.

Abstract: Implementations relate to improved crop field segmentation and crop classification in which boundaries between crop fields are more accurately detected. In various implementations, high-elevation image(s) that capture an area containing multiple demarcated fields may be applied as input across one or more machine learning models to generate a boundary enhancement channel. Each pixel of the boundary enhancement channel may be spatially aligned with a corresponding pixel of the one or more high-elevation images. Moreover, each pixel of the boundary enhancement channel may be classified with a unit angle to a reference location of the field of the multiple demarcated fields that contains the pixel. Based on the boundary enhancement channel, pixel-wise field memberships of pixels of the one or more high-elevation images in the multiple demarcated fields may be determined.

Abstract: A method for ship ballasting includes receiving, at a carbon negative energy storage system, input comprising calcium oxide and water and reacting, within a reaction chamber of the carbon negative energy storage system, the calcium oxide and water to release energy and generate calcium hydroxide. The method includes directing, by the carbon negative energy storage system, the released energy to a requesting end user and providing, by the carbon negative energy storage system, the calcium hydroxide to a marine vessel ballasting system. The method includes releasing a mixture of the calcium hydroxide and ballast water from the marine vessel ballasting system into the ocean to sequester atmospheric CO2.

Abstract: Techniques for processing ore include the steps of causing an imaging capture system to record a plurality of images of a stream of ore fragments en route from a first location in an ore processing facility to a second location in the ore processing facility; correlating the plurality of images of the stream of ore fragments with at least one or more characteristics of the ore fragments using a machine learning model that includes a plurality of ore parameter measurements associated with the one or more characteristics of the ore fragments; determining, based on the correlation, at least one of the one or more characteristics of the ore fragments; and generating, for display on a user computing device, data indicating the one or more characteristics of the ore fragments or data indicating an action or decision based on the one or more characteristics of the ore fragments.

Abstract: An example method may include receiving, from a client computing device, an indication of a target drop-off spot for an object within a first virtual model of a first region of a delivery destination. A second virtual model of a second region of the delivery destination may be determined based on sensor data received from one or more sensors on a delivery vehicle. A mapping may be determined between physical features represented in the first virtual model and physical features represented in the second virtual model to determine an overlapping region between the first and second virtual models. A position of the target drop-off spot within the second virtual model may be determined based on the overlapping region. Based on the position of the target drop-off spot within the second virtual model, the delivery vehicle may be navigated to the target drop-off spot to drop off the object.

Abstract: The present disclosure relates to techniques for deformulating the spectra of arbitrary compound formulations such as polymer formulations into their chemical components. Particularly, aspects of the present disclosure are directed to obtaining an initial set of spectra for a plurality of samples comprising pure samples and composite samples, constructing a basis set of spectra for a plurality of pure samples based on the initial set of spectra, and providing or outputting the basis set of spectrum. The basis set of spectra is constructed in an iterative process that attempts to decompose, using a decomposition algorithm or model, the spectrum from the initial set of spectra in order to differentiate the pure samples from the composite samples. The basis set of spectra may then be used to deduce the composition of a material from a spectrogram.

Abstract: Implementations are described herein for training and using machine learning to determine mappings between matching nodes of graphs representing predecessor source code snippets and graphs representing successor source code snippets. In various implementations, first and second graphs may be obtained, wherein the first graph represents a predecessor source code snippet and the second graph represents a successor source code snippet. The first graph and the second graph may be applied as inputs across a trained machine learning model to generate node similarity measures between individual nodes of the first graph and nodes of the second graph. Based on the node similarity measures, a mapping may be determined across the first and second graphs between pairs of matching nodes.

Abstract: Implementations are described herein are directed to reconciling disparate orientations of multiple vision sensors deployed on a mobile robot (or other mobile vehicle) by altering orientations of the vision sensors or digital images they generate. In various implementations, this reconciliation may be performed with little or no ground truth knowledge of movement of the robot. Techniques described herein also avoid the use of visual indicia of known dimensions and/or other conventional tools for determining vision sensor orientations. Instead, techniques described herein allow vision sensor orientations to be determined and/or reconciled using less resources, and are more scalable than conventional techniques.

Abstract: Systems and methods related to providing configurations of robotic devices are provided. A computing device can receive a configuration request for a robotic device including environmental information and task information for tasks requested to be performed by the robotic device in an environment. The computing device can determine task-associated regions in the environment. A task-associated region for a given task can include a region of the environment that the robotic device is expected to reach while performing the given task. Based at least on the task-associated regions, the computing device can determine respective dimensions of components of the robotic device and an arrangement for assembling the components into the robotic device so that the robotic device is configured to perform at least one task in the environment. The computing device can provide a configuration that includes the respectively determined dimensions and the determined arrangement.

Abstract: Methods and apparatus related to training and/or utilizing a convolutional neural network to generate grasping parameters for an object. The grasping parameters can be used by a robot control system to enable the robot control system to position a robot grasping end effector to grasp the object. The trained convolutional neural network provides a direct regression from image data to grasping parameters. For example, the convolutional neural network may be trained to enable generation of grasping parameters in a single regression through the convolutional neural network. In some implementations, the grasping parameters may define at least: a “reference point” for positioning the grasping end effector for the grasp; and an orientation of the grasping end effector for the grasp.

Abstract: Grasping of an object, by an end effector of a robot, based on a final grasp pose, of the end effector, that is determined after the end effector has been traversed to a pre-grasp pose. An end effector vision component can be utilized to capture instance(s) of end effector vision data after the end effector has been traversed to the pre-grasp pose, and the final grasp pose can be determined based on the end effector vision data. For example, the final grasp pose can be determined based on selecting instance(s) of pre-stored visual features(s) that satisfy similarity condition(s) relative to current visual features of the instance(s) of end effector vision data, and determining the final grasp pose based on pre-stored grasp criteria stored in association with the selected instance(s) of pre-stored visual feature(s).

Abstract: A compound motor-generator system including a first motor-generator and a second motor-generator. The first motor generator includes a stator having a set of three-phase field windings and a first rotor disposed inside and coaxial with the stator and configured to rotate relative to the stator. The second motor-generator includes a rotational stator and a second rotor coupled to a common shaft with the rotor of the first motor-generator and disposed inside and coaxial to the rotational stator. The rotational stator is configured to rotate relative to the second rotor and at a higher rotational speed than the second rotor.

Abstract: A device is provided. The device includes a worm drive comprising a worm and a worm gear. The device also includes an actuator comprising a motor, a shaft, and the worm, wherein the shaft is configured to rotate about a shaft axis, and the actuator is configured to (i) drive the worm drive, and (ii) move linearly along the shaft axis. The device also includes a first spring and a second spring, wherein the second ends are fixed, and wherein the first and second springs are configured to resist movement of the actuator along the shaft axis in opposite directions as a result of forces transmitted through the worm drive. The device further includes a linear position encoder configured to determine a position of the actuator along the shaft axis.

Abstract: Deep machine learning methods and apparatus related to the manipulation of an object by an end effector of a robot are described herein. Some implementations relate to training an action prediction network to predict a probability density which can include candidate actions of successful grasps by the end effector given an input image. Some implementations are directed to utilization of an action prediction network to visually servo a grasping end effector of a robot to achieve a successful grasp of an object by the grasping end effector.

Abstract: Embodiments are provided that include receiving sensor data from a sensor positioned at a plurality of positions in an environment. The environment includes a plurality of landmarks. The embodiments also include determining, based on the sensor data, a subset of the plurality of landmarks detected at each of the plurality of positions. The embodiments further include determining, based on the subset of the plurality of landmarks detected at each of the plurality of positions, a detection frequency of each landmark. The embodiments additionally include determining, based on the determined detection frequency of each landmark, a localization viability metric associated with each landmark. The embodiments still further include providing for display, via a user interface, a map of the environment. The map includes an indication of the localization viability metric associated with each landmark.

Abstract: In embodiments, obtaining a plurality of image sets associated with a geographical region and a time period, wherein each image set of the plurality of image sets comprises multi-spectral and time series images that depict a respective particular portion of the geographical region during the time period, and predicting one or more crop types growing in each of particular locations within the particular portion of the geographical region associated with an image set of the plurality of image sets. Determining a crop type classification for each of the particular locations based on the predicted one or more crop types for the respective particular locations, and generating a crop indicative image comprising at least one image of the multi-spectral and time series images of the image set overlaid with indications of the crop type classification determined for the respective particular locations.

Abstract: Some implementations herein relate to a graphical user interface (GUI) that facilitates dynamically partitioning agricultural fields into clusters on an individual agricultural field-basis using agricultural features. A map of a geographic area containing a plurality of agricultural fields may be rendered as part of a GUI. The agricultural fields may be partitioned into a first set of clusters based on a first granularity value and agricultural features of individual agricultural fields. The individual agricultural fields may be visually annotated in the GUI to convey the first set of clusters of similar agricultural fields. Upon receipt of a second granularity value different from the first granularity value, the agricultural fields may be partitioned into a second set of clusters of similar agricultural fields. The map of the geographic area may be updated so that individual agricultural fields are visually annotated to convey the second set of clusters.

Abstract: In some embodiments, a method of ensuring fabricability of a segmented design for a physical device to be fabricated by a fabrication system is provided. A proposed segmented design is searched for forbidden patterns in a set of forbidden patterns. Segments from the proposed segmented design that appear in forbidden patterns are added to a set of unfabricable segments. A material indicated by at least one unfabricable segment from the set of unfabricable segments is changed to create an updated segmented design, and the updated segmented design is searched for the forbidden patterns in the set of forbidden patterns. In response to determining that the updated segmented design includes at least one forbidden pattern, the adding, changing, and searching actions are repeated. In response to determining that the updated segmented design does not include any forbidden patterns, an indication is generated that the updated segmented design is fabricable.