Abstract: A method and system for optimizing structural parameters of an electromagnetic device is described that includes performing operations. The operations include performing a time-forward simulation of a field response in a simulated environment describing the electromagnetic device and extracting decomposition components from the field response to compute a loss value. The operations further include backpropagating the loss value backwards in time using the decomposition components to determine an influence of changes in the structural parameters of the electromagnetic device on the loss value. The operations further include generating a revised description of the electromagnetic device by updating the structural parameters to reduce the loss value.

Abstract: Methods, apparatus, and computer-readable media for determining and utilizing human corrections to robot actions. In some implementations, in response to determining a human correction of a robot action, a correction instance is generated that includes sensor data, captured by one or more sensors of the robot, that is relevant to the corrected action. The correction instance can further include determined incorrect parameter(s) utilized in performing the robot action and/or correction information that is based on the human correction. The correction instance can be utilized to generate training example(s) for training one or model(s), such as neural network model(s), corresponding to those used in determining the incorrect parameter(s). In various implementations, the training is based on correction instances from multiple robots. After a revised version of a model is generated, the revised version can thereafter be utilized by one or more of the multiple robots.

Abstract: Methods, systems, and apparatus, including computer-readable storage devices, for enhanced object discrimination by robots are described. In some implementations, a robot obtains image data from a camera of the robot, where the image data indicates a detected object. The robot classifies the object as having a particular object type based on the image data. The robot evaluates a current state of the object that includes a predetermined set of object state characteristics. The robot determines that a software application of the robot has indicated (i) an object type and (ii) one or more object state criteria that satisfy the current state of the detected object. The robot provides to the software application data indicating detection of the object and a position of the object.

Abstract: Methods, systems, and apparatus for performing convolutional computations using an optical system. In some aspects computations for a neural network are performed in a digital domain using electronic circuitry, the neural network including a convolutional layer. Input data for the convolutional layer of the neural network is obtained, and a convolution or correlation computation on the input data in an analog domain using an optical correlator module is performed to generate an optical correlator module output. Based on the optical correlator module output, data is processed through additional layers of the neural network in the digital domain using the electronic circuitry.

Abstract: Methods, apparatus, and computer readable media that are related to 3D object detection and pose determination and that may optionally increase the robustness and/or efficiency of the 3D object recognition and pose determination. Some implementations are generally directed to techniques for generating an object model of an object based on model point cloud data of the object. Some implementations of the present disclosure are additionally and/or alternatively directed to techniques for application of acquired 3D scene point cloud data to a stored object model of an object to detect the object and/or determine the pose of the object.

Abstract: Embodiments of techniques for inverse design of physical devices are described herein, in the context of generating designs for photonic integrated circuits (including a multi-channel photonic demultiplexer). In some embodiments, an initial design of the physical device is received, and a plurality of sets of operating conditions for fabrication of the physical device are determined. In some embodiments, the performance of the physical device as fabricated under the sets of operating conditions is simulated, and a total performance loss value is backpropagated to determine a gradient to be used to update the initial design. In some embodiments, instead of simulating fabrication of the physical device under the sets of operating conditions, a robustness loss is determined and combined with the performance loss to determine the gradient.

Abstract: Methods and apparatus related to generating a model for an object encountered by a robot in its environment, where the object is one that the robot is unable to recognize utilizing existing models associated with the robot. The model is generated based on vision sensor data that captures the object from multiple vantages and that is captured by a vision sensor associated with the robot, such as a vision sensor coupled to the robot. The model may be provided for use by the robot in detecting the object and/or for use in estimating the pose of the object.

Abstract: Methods, systems, and apparatus for receiving a mapping of a property that includes a three-dimensional representation of the property, receiving observations of the property that each depict a portion of the property, providing the mapping and the observations to an object mapping engine, receiving an object mapping of the property, wherein the object mapping includes a plurality of object labels that each identify an object that was recognized from the observations and a location of the object within the three-dimensional representation that corresponds to a physical location of the object in the property, and obtaining a semantic mapping of the property that identifies semantic zones of the property with respect to the three-dimensional representation, wherein the semantic mapping is generated based on an output that results from a semantic mapping model processing the object mapping.

Abstract: Example embodiments are directed to systems and methods for denying power to unauthorized power loads within a spherical waveguide bounded by the Earth's surface. One or more eigenmodes of the Earth-ionosphere waveguide may be computed based on a mathematical model incorporating electrical properties of the terrestrial surface and plasma physics of the ionosphere. A phased array of wave guide couplers, each including an array of electrically-connected waveguide-coupling elements, may be configured at different geographic locations for coupling to the one or more eigenmodes and generating standing waves in the Earth-ionosphere waveguide. Power loads may be detected via power reflections, and power nodes and nulls of the standing waves may be steered with respect to specified locations by adjusting relative phases and/or amplitudes of the waveguide couplers. Locations of power loads may be triangulating reflections. Power nulls of the standing waves may be steered to locations of unauthorized power loads.

Abstract: A two-channel photonic demultiplexer includes an input region to receive a multi-channel optical signal, two output regions, each adapted to receive a corresponding one of two distinct wavelength channels demultiplexed from the multi-channel optical signal, and a dispersive region including a first material and a second material inhomogeneously interspersed to form a plurality of interfaces that collectively structure the dispersive region to optically separate each of the two distinct wavelength channels from the multi-channel optical signal and respectively guide the first distinct wavelength channel to a first output region and the second distinct wavelength channel to the second output region when the input region receives the multi-channel optical signal. At least one of the first material or the second material is structured within the dispersive region to be schematically reproducible by a feature shape with a pre-determined width.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, of training a global policy neural network. One of the methods includes initializing an instance of the robotic task for multiple local workers, generating a trajectory of state-action pairs by selecting actions to be performed by the robotic agent while performing the instance of the robotic task, optimizing a local policy controller on the trajectory, generating an optimized trajectory using the optimized local controller, and storing the optimized trajectory in a replay memory associated with the local worker. The method includes sampling, for multiple global workers, an optimized trajectory from one of one or more replay memories associated with the global worker, and training the replica of the global policy neural network maintained by the global worker on the sampled optimized trajectory to determine delta values for the parameters of the global policy neural network.

Abstract: A method for simulating a biochemical environment utilizes a heterogeneous process model, which evaluates both a flux balance analysis and one or more detailed models each on a different but overlapping sets of molecules in the biochemical environment. The heterogeneous process model evaluates the flux balance analysis based on a stoichiometric matrix, a flux vector including initial internal exchange flux values, and an objective function. The heterogeneous process model evaluates the one or more detailed models based on an initial set of molecule concentrations and a plurality of detailed model parameters. The results are then used to update the exchange fluxes and molecules concentrations. The process is repeated thereby integrating the results of the flux balance analysis with the one or more detailed models.

Abstract: An apparatus comprising a spherical vessel, where the spherical vessel further includes a wall, where the wall is formed from one or more layers of material and a gas, where the gas may be lighter-than-air and is enclosed by the wall.

Abstract: Implementations are described herein for generating embeddings of source code using both the language and graph domains, and leveraging combinations of these semantically-rich and structurally-informative embeddings for various purposes. In various implementations, tokens of a source code snippet may be applied as input across a sequence-processing machine learning model to generate a plurality of token embeddings. A graph may also be generated based on the source code snippet. A joint representation may be generated based on the graph and the incorporated token embeddings. The joint representation generated from the source code snippet may be compared to one or more other joint representations generated from one or more other source code snippets to make a determination about the source code snippet.

Abstract: Methods, apparatus, systems, and computer-readable media are provided for delegating object type and/or pose detection to a plurality of “targeted object recognition modules.” In some implementations, a method may be provided that includes: operating an object recognition client to facilitate object recognition for a robot; receiving, by the object recognition client, sensor data indicative of an observed object in an environment; providing, by the object recognition client, to each of a plurality of remotely-hosted targeted object recognition modules, data indicative of the observed object; receiving, by the object recognition client, from one or more of the plurality of targeted object recognition modules, one or more inferences about an object type or pose of the observed object; and determining, by the object recognition client, information about the observed object, such as its object type and/or pose, based on the one or more inferences.

Abstract: An optical communication system includes an optical transmitter and one or more processors. The optical transmitter is configured to output an optical signal, and includes an average-power-limited optical amplifier, such as an erbium-doped fiber amplifier (EDFA). The one or more processors are configured to receive optical signal data related to a received power for a communication link from a remote communication system and determine that the optical signal data is likely to fall below a minimum received power within a time interval. In response to the determination, the one or more processors are configured to determine a duty cycle of the optical transmitter based on a minimum on-cycle length and a predicted EDFA output power and operate the optical transmitter using the determined duty cycle to transmit an on-cycle power that is no less than the minimum required receiver power for error-free operation of the communication link.

Abstract: Techniques for ensuring fabricability of segmented designs are provided. In some embodiments, a computing system receives a design specification to be used to create a proposed segmented design. Each segment of the proposed segmented design indicates a material within the segment. The proposed segmented design includes segments that indicate at least two different materials. The computing system chooses a segment to touch with a paintbrush pattern for a first material. The computing system updates a set of material statuses for each material based on the chosen segment. The computing system updates a set of touch statuses for each material based on the chosen segment. The choosing and updating actions are repeated until all segments are associated with a material. The computing system generates the proposed segmented design based on the sets of material statuses.

Abstract: In some embodiments, techniques are disclosed for creating a fabricable segmented design for a physical device for which a loss is to be minimized. A computing system determines costs associated with touching each segment of a proposed segmented design based on a design specification with a first paintbrush pattern associated with a first material or a second paintbrush pattern associated with a second material. The computing system chooses a segment to touch with the first paintbrush pattern or the second paintbrush pattern based on the costs, and touches the segment with the pattern. These actions are repeated until all segments have been associated with the first material or the second material. The computing system estimates a gradient of the loss, and backpropagates the gradient of the loss to update the design specification.

Abstract: A method includes receiving input data indicative of tactile input at a bidirectional jog control device associated with a seven-degree-of-freedom (7DOF) robotic arm, where the 7DOF robotic arm is in a first arm configuration with an end effector of the 7DOF robotic arm positioned at a first pose in an environment. Based on the input data, the method further includes determining a direction to jog the 7DOF robotic arm through a null space while keeping the end effector fixed at the first pose in the environment. The method additionally includes controlling the 7DOF robotic arm to jog through the null space in the determined direction to a second arm configuration, where the end effector is positioned at the first pose in the environment when the 7DOF robotic arm is in the second arm configuration.