Abstract: System and method for end-to-end differentiable joint image refinement and perception are provided. A learning machine employs an image acquisition device for acquiring a set of training raw images. A processor determines a representation of a raw image, initializes a set of image representation parameters, defines a set of analysis parameters of an image analysis network configured to process the image's representation, and jointly trains the set of representation parameters and the set of analysis parameters to optimize a combined objective function. A module for transforming pixel-values of the raw image to produce a transformed image comprising pixels of variance-stabilized values, a module for successively performing processes of soft camera projection and image projection, and a module for inverse transforming the transformed pixels are disclosed. The image projection performs multi-level spatial convolution, pooling, subsampling, and interpolation.

Abstract: A method includes, storing a set of valid codewords including: a first valid functional codeword representing a functional timeout state of a second controller; a first valid fault codeword representing a fault timeout state of the second controller and characterized by a minimum hamming distance from the first valid functional codeword; a second valid functional codeword representing a functional state of a system; and a second valid fault codeword representing a fault state of the system; in response to detecting receipt of a safety message from the second controller within a predefined time quantum, storing the first valid functional codeword in a first memory; in response to detecting a match between contents of the first memory and the first valid functional codeword, outputting the second valid functional codeword; in response to detecting a mismatch between contents of the first memory and every codeword in the first set of valid codewords, outputting the second valid fault codeword.

Abstract: Systems and methods for vehicle controllers for agricultural and industrial applications are described. For example, a method includes accessing a map data structure storing a map representing locations of physical objects in a geographic area; accessing current point cloud data captured using a distance sensor connected to a vehicle; detecting a crop row based on the current point cloud data; matching the detected crop row with a crop row represented in the map; determining an estimate of a current location of the vehicle based on a current position in relation to the detected crop row; and controlling one or more actuators to cause the vehicle to move from the current location of the vehicle to a target location.

Abstract: Disclosed is a system and associated methods for optimizing order fulfillment. The system determines the containers in a site with items for a set of received orders, and reserves different slot types of an item cache for each container based on the properties of the contained items or container. The system transfers a set of containers to the item cache by placing each of the set of containers into a slot that is associated with a slot type that matches a slot type reserved for that container. The system replaces a first container from the set of containers upon delivering a second container to the item cache and other containers from the set of containers in the slot type reserved for the second container having items for unfulfilled orders and the first container not having items for unfulfilled orders once the second container is brought to the item cache.

Abstract: A mobile robot is configured for operation in a commercial or industrial setting, such as an office building or retail store. The robot can patrol one or more routes within a building, and can detect violations of security policies by objects, building infrastructure and security systems, or individuals. In response to the detected violations, the robot can perform one or more security operations. The robot can include a removable fabric panel, enabling sensors within the robot body to capture signals that propagate through the fabric. In addition, the robot can scan RFID tags of objects within an area, for instance coupled to store inventory. Likewise, the robot can generate or update one or more semantic maps for use by the robot in navigating an area and for measuring compliance with security policies.

Abstract: A system includes wearable devices positioned on a subject in different locations. Each wearable device includes motion sensors that measure the subject's movement in three dimensions. The motion sensors generate raw sensory data as the subject performs a physical movement. A data filter is selected based on a condition of the subject and a designated movement corresponding to the physical movement, and used to convert the raw sensory data into formatted data. A level of compliance of the physical movement with a movement model for the designated movement is determined by applying comparative modeling techniques to the formatted data and the movement model. Real-time feedback is delivered dynamically to the subject by the wearable devices during the performance of the physical movement based on the level of compliance. The movement model can be generated, and the comparative modeling techniques can be selected, based on the condition of the subject.

Abstract: Embodiments are directed providing a stage studio for immersive 3-D video capture. A two-dimensional video of a scene may be captured with frame cameras oriented towards a center of the scene. Paths may be scanned across objects in the scene with signal beams oriented towards the center of the scene. Events may be generated based on signal beams that are reflected by the objects and detected by event cameras oriented towards the center of the scene. Trajectories may be generated based on the paths and the events. A scene that includes a representation of the objects based on the captured two-dimensional video and the trajectories may be generated such that a position and an orientation of the represented objects in the scene are based on the trajectories.

Abstract: Systems and methods for agricultural lane following are described. For example, a method includes accessing range data captured using a distance sensor connected to a vehicle and/or image data captured using an image sensor connected to a vehicle; detecting a crop row based on the range data and/or the image data to obtain position data for the crop row; determining, based on the position data for the crop row, a yaw and a lateral position of the vehicle with respect to a lane bounded by the crop row; and based on the yaw and the lateral position, controlling the vehicle to move along a length of the lane bounded by the crop row.

Abstract: System and method for joint refinement and perception of images are provided. A learning machine employs an image acquisition device for acquiring a set of training raw images. A processor determines a representation of a raw image, initializes a set of image representation parameters, defines a set of analysis parameters of an image analysis network configured to process the image's representation, and jointly trains the set of representation parameters and the set of analysis parameters to optimize a combined objective function. A module for transforming pixel-values of the raw image to produce a transformed image comprising pixels of variance-stabilized values, a module for successively performing processes of soft camera projection and image projection, and a module for inverse transforming the transformed pixels are disclosed. The image projection performs multi-level spatial convolution, pooling, subsampling, and interpolation.

Abstract: A robot system useful in manufacturing environments for diverse applications. The system is characterized by an elongate robot arm operable to selectively position an end effector. The robot arm is configured for movement by a CPU based control system and/or by physical manipulation by a human operator.

Abstract: A system for control of a mobility device comprising a controller for analyzing data from at least one sensor on the mobility device, wherein the data is used to determine the gait of user. The gait data is then used to provide motion command to an electric motor on the mobility device.

Abstract: A method for controlling movement of a robot includes inputting a first linear velocity parameter and a first angular velocity parameter, which are specified as the robot moves, into a first limit model for limiting a centripetal acceleration to correct the first linear velocity parameter and the first angular velocity parameter, thereby calculating a second linear velocity parameter and a second angular velocity parameter, inputting the second linear velocity parameter and the second angular velocity parameter into at least one of a second limit model for limiting a linear velocity and a third limit model for limiting an angular velocity to correct the second linear velocity parameter and the second angular velocity parameter, thereby calculating a third linear velocity parameter and a third angular velocity parameter, and controlling the movement of the robot based on the third linear velocity parameter and the third angular velocity parameter.

Abstract: Systems and techniques for detecting a difference in orientation between a lawn mower and a surface on which the mower is mowing, or between portions of such surface, based on one or more of sensor data or map data are discussed. The difference in orientation may then be used to control the lawn mower by, for example, raising a deck height or idling a blade speed, though other actuations are contemplated. In some examples, an amount the deck height is raised may be based on the difference in orientation. Based on a subsequent difference detected being at or below a threshold difference, the mower may return to previous operating conditions including, but not limited to, returning a blade speed to a previous blade speed and/or returning the deck height to a previous deck height, among other controls.

Abstract: An apparatuses, methods and systems for a desktop occupancy sensing device are disclosed. One desktop apparatus includes an occupancy sensor, a charging interface, a controller, and a structure. The occupancy sensor operates to sense occupancy proximate to the desktop apparatus. The controller operates to determine occupancy of a desktop based on the sensed occupation, and the controller operates to communicate the determined occupancy to an upstream network. The occupancy sensor, the charging interface, and an interface to the upstream network are disposed within the structure as a single unit, and the structure is adapted to be placed on or affixed to the desktop.

Abstract: An order fulfillment system controller receives orders for different objects, generates a number of batched tasks that is less than the number of received orders, and determines a set of the orders with a total distance between them that is greater than a total distance between other sets of the orders. The controller allocates a different order from the set of order to a different batched task, and allocates each unallocated order to a selected batched task based on the distance between objects of the unallocated order and objects previously allocated to the selected batched task being less than the distance between objects of the unallocated order and objects previously allocated to other batched tasks. The controller controls different agents according to the allocation of different subsets of orders to different batched tasks.

Abstract: A system has a virtual-world (VW) controller and a physical-world (PW) controller. The pairing of a PW element with a VW element establishes them as corresponding physical and virtual twins. The VW controller and/or the PW controller receives measurements from one or more sensors characterizing aspects of the physical world, the VW controller generates the virtual twin, and the VW controller and/or the PW controller generates commands for one or more actuators affecting aspects of the physical world. To coordinate the corresponding virtual and physical twins, (i) the VW controller controls the virtual twin based on the physical twin or (ii) the PW controller controls the physical twin based on the virtual twin. Depending on the operating mode, one of the VW and PW controllers is a master controller, and the other is a slave controller, where the virtual and physical twins are both controlled based on one of VW or PW forces.

Abstract: A method for controlling a robot is provided. The method includes the steps of: determining a first comparison axis with reference to a first target area specified by a camera module of a robot, and determining a second comparison axis with reference to a second target area specified by a scanner module of the robot and associated with the first target area; and correcting a reference coordinate system associated with the camera module with reference to a relationship between the first comparison axis and the second comparison axis.

Abstract: Methods, systems and apparatus for providing a notification of a wet floor are provided. A scrubbing robot for scrubbing a floor and having one or more liquid carrying components has a sensor coupled to it positioned for collecting data about an area of the floor proximate to the scrubbing robot and along a path over which the robot has travelled. An application is stored on the computer for determining that liquid is on the floor by analysing the data for a presence of a second set of parameters corresponding to or crossing (from above to below or from below to above) a set of threshold parameters that is indicative of the presence of liquid on the floor. A notification module coupled to the scrubbing robot issues at least one of a human detectable and a computer detectable notification in response to the application determining that liquid is on the floor.

Abstract: A multi-level induction station is disclosed. The multi-level induction station includes multiple sorting levels arranged in a vertical stacked configuration; a vertically traveling automated lift that carries articles vertically between the sorting levels along a frame positioned proximate to the multiple sorting levels; an information acquisition device positioned to read information from an article on the lift; multiple transport devices positioned on the sorting levels to receive the articles; and a control system that directs the multi-level induction station to determine a destination for an article and direct the lift to a platform where a transport device is at an article-receiving position at the sorting level to receive the article.

Abstract: A patient console for a robotic surgical system can include a base, a vertical lift attached to a top of the base and configured to provide up and down motion in a vertical axis, a yaw rotation device attached to the top of the vertical lift and configured to provide a yaw rotation about the vertical axis, a pitch rotation device attached to the top of the yaw rotation device and configured to provide a pitch rotation about a pitch axis orthogonal to the vertical axis, a translation device attached to the top of the pitch rotation device and configured to provide sliding translation along a translation axis, and a roll rotation device attached to the translation device to roll relative to the translation device about a roll axis to provide a roll to an instrument controller assembly. An angle of the translation axis and the roll axis relative to horizontal can be a function of the pitch rotation provided by the pitch rotation device.