Abstract: Techniques described herein relate to using reduced-dimensionality embeddings generated from robot sensor data to identify predetermined semantic labels that guide robot interaction with objects. In various implementations, sensor data obtained from one or more sensors of a robot includes data indicative of an object observed in an environment in which the robot operates. The sensor data is processed utilizing a first trained machine learning model to generate a first embedded feature vector that maps the data indicative of the object to an embedding space. Nearest neighbor(s) of the first embedded feature vector is identified in the embedding space. Semantic label(s) are identified based on the nearest neighbor(s). A given grasp option is selected from enumerated grasp options previously associated with the semantic label(s). The robot is operated to interact with the object based on the pose and using the given grasp option.

Abstract: An example system includes a vehicle and a sensor connected to the vehicle. The system may receive a predetermined path for the vehicle to follow. The system may also receive a plurality of objectives, associated with a corresponding set of sensor data, for which to collect sensor data. The system may determine, for each of the plurality of objectives, a portion of the environment for the sensor to scan to acquire the corresponding set of sensor data. The system may determine, based on the portion of the environment determined for each of the plurality of objectives, a sensor trajectory through which to move the sensor. The system may cause the sensor to move through the determined sensor trajectory and scan portions of the environment corresponding to the determined sensor trajectory as the vehicle moves along the predetermined path.

Abstract: Methods, apparatus, and computer readable media related to combining and/or training one or more neural network modules based on version identifier(s) assigned to the neural network module(s). Some implementations are directed to using version identifiers of neural network modules in determining whether and/or how to combine multiple neural network modules to generate a combined neural network model for use by a robot and/or other apparatus. Some implementations are additionally or alternatively directed to assigning a version identifier to an endpoint of a neural network module based on one or more other neural network modules to which the neural network module is joined during training of the neural network module.

Abstract: Systems and methods for detecting and locating power loads within a spherical waveguide bounded by the Earth's surface are disclosed. 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 ionospheric layer. 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 by way of power reflections, and by adjusting relative phases and/or amplitudes of the waveguide couplers, as well as frequencies, power nodes and nulls of the standing waves may be steered with respect to specified locations. Using reflections and steering, locations of power loads may be triangulated.

Abstract: Methods, apparatus, systems, and computer-readable media are provided for visually annotating rendered multi-dimensional representations of robot environments. In various implementations, an entity may be identified that is present with a telepresence robot in an environment. A measure of potential interest of a user in the entity may be calculated based on a record of one or more interactions between the user and one or more computing devices. In some implementations, the one or more interactions may be for purposes other than directly operating the telepresence robot. In various implementations, a multi-dimensional representation of the environment may be rendered as part of a graphical user interface operable by the user to control the telepresence robot. In various implementations, a visual annotation may be selectively rendered within the multi-dimensional representation of the environment in association with the entity based on the measure of potential interest.

Abstract: Aspects of the disclosure provide for a controller of a system. The controller is configured to apply a drive to transition the system from a first state to a second state and to continually adjust the drive in response to a detected disturbance quantity in real-time by performing a loop until the system is in the second state. The loop comprises determining a predicted state of the system at a given instance based on the obtained current state of the system and the applied drive, obtaining the current state of the system at the given instance using one or more sensors, determining a disturbance quantity amount based on a difference between the predicted state at the given instance and an actual state of the system at the given instance, determining an adjustment to the drive based on the disturbance quantity amount, and applying the adjusted drive to the system.

Abstract: An example device includes a rigid plate, an inner element, a plurality of connecting flexural elements coupled between the inner element and rigid plate, and a hardstop that extends through the inner element and couples to the rigid plate. The inner element has a plurality of reflective surface areas that are configured to reflect light to a sensor. The plurality of connecting flexural elements allow the inner element to move relative to rigid plate. The hardstop contacts the inner element when a load applied on the device exceeds a threshold load.

Abstract: Methods, systems, and apparatus, including computer-readable media storing executable instructions, for enhancing robot learning. In some implementations, a robot stores first embeddings generated using a first machine learning model, and the first embeddings include one or more first private embeddings that are not shared with other robots. The robot receives a second machine learning model from a server system over a communication network. The robot generates a second private embedding for each of the one or more first private embeddings using the second machine learning model. The robot adds the second private embeddings to the cache of the robot and removes the one or more first private embeddings from the cache of the robot.

Abstract: An example device includes an inner element, an outer surrounding element, and a plurality of connecting flexural elements coupled between the inner element and the outer surrounding element. The inner element has a plurality of reflective surface areas that are configured to reflect light to a sensor. The outer surrounding element surrounds the inner element. The plurality of connecting flexural elements allow the inner element to move relative to the outer surrounding element.

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: A bipartite graph structure is utilized to better store data. The bipartite graph structure may be used in a biochemical database to efficiently store a variety of molecules and processes that might occur between the molecules. Molecules are represented as molecule nodes, which may have metadata fields including a molecule name, a molecule type, a molecular formula, a sequence, a molecular charge, a set of molecular properties, and a set of component molecules. Processes operating on the molecules are represented by process nodes, which may have metadata fields including a process name, a set of process roles, a set of process properties, and a set of sub-processes. Edges, called roles, each associate a molecule node with a process node and represent the role the associated molecule plays in the associated process. The roles may contain metadata identifying the role type and the stoichiometry coefficient of the molecule in the process.

Abstract: An electroencephalogram (EEG) system is disclosed. The EEG system includes an EEG controller and an EEG sensor that includes a contact surface, the contact surface including an electrically-conductive portion in communication with the EEG controller and a sensor release element in communication with the EEG controller, the sensor release element being configured to perform, in response to a signal from the EEG controller, a release action to reduce adhesion of an electrically-conductive gel between the contact of the sensor and the user's skin.

Abstract: A method for obtaining an electroencephalogram (EEG) of a user is disclosed. A reference sensor is attached to the user by connecting a first component of the reference sensor to a second component of the reference sensor, at least a portion of the first component being sub-dermally implanted on or adjacent to a mastoid process of the user. At least one active sensor is attached to the user. A first signal is detected from the reference sensor simultaneously as a second signal is detected from the at least one active sensor. The EEG is obtained based on the first signal and the second signal.

Abstract: An in-ear device includes a housing shaped to hold the in-ear device in an ear, and an audio package, disposed in the housing, to emit sound. A tympanic membrane measurement unit (TMMU) is structured to measure a movement of a tympanic membrane in the ear caused by external sound received by the tympanic membrane, and a controller is coupled to the audio package and the TMMU. The controller includes logic that when executed by the controller causes the in-ear device to perform operations. The operations include measuring a movement of the tympanic membrane, and in response to measuring the movement of the tympanic membrane, outputting sound from the audio package to destructively interfere with the external sound received by the tympanic membrane.

Abstract: A photovoltaic power source includes a receptacle to receive a photofuel including a liquid, and one or more photovoltaic cells positioned within the receptacle to receive light emitted from the photofuel when the photofuel is in the receptacle. The photovoltaic power source also includes power circuitry coupled to the one or more photovoltaic cells to receive a photocurrent generated by the one or more photovoltaic cells when the one or more photovoltaic cells receive the light emitted from the photofuel. In response to the photocurrent, the power circuitry is coupled to output electricity.

Abstract: Embodiments are provided that include a method that includes causing a mobile robot to navigate toward a standing test device placed within an environment of the mobile robot. The method also includes receiving, from a proximity sensor on the standing test device, sensor data indicative of proximity of the mobile robot to the standing test device as the mobile robot navigates toward the standing test device. The method further includes determining, based on the sensor data received from the proximity sensor on the standing test device, an approach motion profile followed by the mobile robot in navigating toward the standing device. The method additionally includes providing a control instruction for the mobile robot based on the determined approach motion profile.

Abstract: A laser tracking system for determining pose information of a rigid object is disclosed. The laser tracking system includes three or more retroreflectors, three or more sets of multiple laser trackers, and an electronic controller. Each retroreflector is secured to the rigid object that is moveable within a frame of reference. For each set of laser trackers, each laser tracker is configured to direct a laser beam to and receive a reflected laser beam from an associated one of the retroreflectors within the frame of reference. The electronic controller is in communication with each of the laser trackers and determines the pose information of the rigid object in the reference frame based on information about the fixed location of each laser tracker in the frame of reference and information about a distance of each retroreflector from each laser tracker of the set of laser trackers associated with the retroreflector.

Abstract: Aspects of the disclosure provide techniques for automatic repeat request (ARQ) in a free-space optical communication (FSOC) architecture. These techniques, including block-selective ARQ, adaptive retransmission delay, and random seed scrambling, can be used individually or in combination to combat problems involving frame loss or corruption. These techniques enable the system to rapidly recover by streamlining the retransmission process. For instance, block-selective ARQ acknowledges variable length blocks of frames in the return stream from the receiver to the transmitter. Adaptive retransmission delay allows the retransmission delay to grow in the absence of feedback by the receiver, up to some defined limit. And with random seed sampling, a scrambling sequence is incorporated to aid with frame syncing, which avoids the need for a line code. These aspects of the technology provide a robust communication process, and also reduce overhead costs associated with unnecessary retransmissions.

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.