Abstract: Disclosed in the present disclosure are a transfer reinforcement learning method and apparatus, multi-task reinforcement learning method and apparatus, relating to the field of intelligent control technology. The transfer reinforcement learning method includes determining operational instructions for instructing an agent to perform a first task; determining an inclusion relation between multiple second tasks and the first tasks based on the operational instructions; determining a shared parameter set corresponding to the multiple second tasks based on the inclusion relation between the multiple second tasks and the first task, wherein the shared parameter set includes a plurality of parameters shared by the multiple second tasks; and performing transfer reinforcement learning based on the shared parameter set and the first task to obtain model parameters of a target policy model corresponding to the first task.

Abstract: Aspects of this technical solution can include identifying, by a processor coupled to non-transitory memory, a plurality of bounding boxes for one or more objects depicted in each image of a sequence of images captured during operation of an autonomous vehicle, allocating, by the processor and based on corresponding positions of the bounding boxes in each image and corresponding time stamps, one or more of the bounding boxes to one or more tracking identifiers each indicating trajectories of corresponding objects, generating, by the processor and based on the time stamps and the bounding boxes allocated to each of the tracking identifiers, one or more tracking images for each of the tracking identifiers, each of the tracking images including visual indications of the time stamps, and training, by the processor and based on the tracking images, an artificial intelligence model to output an indication of a type of trajectory.

Abstract: A method is disclosed for reducing impact forces to legged robots as a result of traversing a terrain. The method includes employing actuators and compliant elements to use the contact of a first portion of a foot assembly with a terrain during a step to reduce the vertical velocity of a subsequent portion of the foot assembly so that the vertical velocity of the subsequent portion as it touches the terrain is substantially zero relative to the terrain. Foot assemblies that employ these methods are also provided.

Abstract: A method comprising: accessing a response mapping defining a set of safety-critical functions associated with a safety-critical latency threshold and a set of safety responses, each safety response corresponding to a safety-critical function; executing a time-synchronization protocol with a transmitting system to calculate a clock reference; accessing a safety message schedule indicating an expected arrival time for each safety message in a series of safety messages based on the clock reference; for each safety message in the series of safety messages, calculating a latency of the safety message based on an arrival time of the safety message and the expected arrival time; and in response to a latency of a current safety message in the series of safety messages exceeding the safety-critical latency threshold, initiating the safety response corresponding to the safety-critical function for each safety-critical function in the set of safety-critical functions.

Abstract: An autonomous off-road vehicle (AOV) includes a vehicle body and a vehicle arm coupled to the vehicle body and including a set of prongs and a clamp extending outward from an end of the vehicle arm. A controller causes the vehicle arm to pick up a pile by: aligning the set of prongs with an end of the pile such that a first prong aligns with an opening on a first side of a web of the pile and a second prong aligns with a space on a second side of the web, inserting the set of prongs into the opening on the first side and the space on the second side, and compressing the clamp into an exterior surface of a flange of the pile such that the set of prongs are reciprocally compressed into an interior surface of the flange on either side of the web.

Abstract: Data is received that includes a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system. Thereafter, it is detected whether there is an object within each image. Based on this detection, images in which each object is detected that meet predefined object representation parameters are identified (on an object-by-object basis, etc.). The identified images are provided to a consuming application or process for quality assurance analysis. Related apparatus, systems, techniques and articles are also described.

Abstract: A system includes a group of transceiver nodes and diagnostic circuitry. A first transceiver node of the group broadcasts one or more beacons which a second transceiver node of the group attempts to receive. The second transceiver node provides a reception indication to the diagnostic circuitry. Based on the reception indication, the diagnostic circuitry determines a radio performance change for the first transceiver node and/or second transceiver node. Using a change threshold based on a distribution of radio performance changes for the group, the diagnostic circuitry may determine whether to generate a change indication for the first transceiver node and/or second transceiver node.

Abstract: Aspects of this technical solution can include detecting, by a sensor of a vehicle in a physical environment via visible light, a first object in the physical environment, detecting, by the sensor via the visible light, a first feature a having a digital encoding and located at a surface of the first object, decoding, by a processor of the vehicle and based on the digital encoding, the first feature into a first indication of location corresponding to the first object, generating, by the processor of the vehicle during movement of the vehicle through the physical environment and based on the first indication of location, a location metric corresponding to the vehicle, and modifying, by the processor of the vehicle based on the location metric, operation of the vehicle to navigate the vehicle through the physical environment according to the location metric.

Abstract: Disclosed herein are systems, methods, and apparatuses for calibrating sensors of automated vehicles using calibration targets. A bracket holds the calibration targets and attaches to the automated vehicle proximate to a sensor being calibrated. For some sensors, such as GNSS antennas or similar device for receiving location data from a GNSS, a two-target bracket is fixed at the top of the automated vehicle, nearby the GNSS antenna. For IMUs or similar sensors, a three-target bracket is fixed at location of the automated vehicle proximate to the particular IMU, such as a passenger cabin or on the chassis of the automated vehicle. The calibration targets include a retroreflective surface that reflect signals, such as infrared signals, back to a theodolite (or total station). The theodolite or computer includes preprogrammed offset values indicating the relative positions of the sensor being calibrated and each of the calibration targets of the bracket.

Abstract: A method includes, storing a set of valid codewords including: a first valid functional codeword representing a functional state of a controller subsystem; a first valid fault codeword representing a fault state of the controller subsystem and characterized by a minimum hamming distance from the first valid functional codeword; a second valid functional codeword representing a functional state of a controller; and a second valid fault codeword representing a fault state of the controller; in response to detecting functional operation of the controller subsystem, 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 vision-based position and orientation determination for endovascular tools are disclosed. In one example, a method includes receiving a two-dimensional medical image including a view of at least a distal portion of a medical instrument, the distal portion of the medical instrument including one or more fiducials positioned thereon, the one or more fiducials being radio-opaque and visible in the medical image. The method also includes detecting, within the medical image, a two-dimensional appearance of the one or more fiducials, and based on the two-dimensional appearance of the one or more fiducials, determining at least one of a roll angle of the distal portion of the medical instrument, and an incline of the distal portion of the medical instrument.

Abstract: A method and system for registering a 3D sensor with an autonomous manipulator is provided. The 3D sensor has a field of view and a sensor coordinate system. The autonomous manipulator is a vision-guided manipulator having a work envelope and a manipulator coordinate system. The method includes moving a registration target relative to the sensor in the field of view of the sensor in the work envelope to obtain a plurality of depth maps or images of the target. The depth maps or images are processed to obtain a plurality of extrinsic registration parameters between the manipulator and the sensor.

Abstract: Systems and methods of manipulating/controlling robots. In many scenarios, data collected by a sensor (connected to a robot) may not have very high precision (e.g., a regular commercial/inexpensive sensor) or may be subjected to dynamic environmental changes. Thus, the data collected by the sensor may not indicate the parameter captured by the sensor with high accuracy. The present robotic control system is directed at such scenarios. In some embodiments, the disclosed embodiments can be used for computing a sliding velocity limit boundary for a spatial controller. In some embodiments, the disclosed embodiments can be used for teleoperation of a vehicle located in the field of view of a camera.

Abstract: The present invention provides a medical robot which is hygienic and capable of precisely transferring a catheter, a driving device mounted on the medical robot, and a roller module mounted on the driving device. The roller module includes: a driving unit; and a roller unit rotated about the vertical axis by the driving unit, wherein the roller unit includes at least one roller and at least one shaft independently coupled to the at least one roller of the roller unit and rotated by the driving unit, wherein at least one groove is formed on the at least one roller of the roller unit and at least a part of the at least one shaft of the roller unit is disposed on the at least one groove.

Abstract: A deployment assembly is mounted in a delivery vehicle and holds an autonomous robot that can be deployed to deliver or pick up packages. The assembly has a base and extendable members that move the robot from a retracted position inside the vehicle and an extended position outside the vehicle. The assembly also has securement features that hold the robot in during transport and can also recharge the robot if necessary.

Abstract: Robotic picking devices and methods for performing a picking operation. The methods described herein may involve determining that a picking device is unable to grasp an item and then performing, using a perturbation mechanism, a perturbation operation to perturb the item so that the picking device is more likely to grasp the item by executing a subsequent grasp attempt.

Abstract: An exoskeleton system, the exoskeleton system comprising one or more actuator units that include a fluidic actuator, one or more sensors and an exoskeleton device. The exoskeleton device includes a fluidic system, and a processor and memory, the memory storing instructions, that when executed by the processor, are configured to control the exoskeleton system to introduce fluid to the fluidic actuator of the one or more actuator units to cause actuation of the fluidic actuator of the one or more actuator units. The exoskeleton system may be configured to operate in, on or around a body of water and can be water and/or corrosion resistant.

Abstract: One aspect provides a modular inspection robot for inspecting vertical shafts, chambers or tunnels. An embodiment provides related methods and products. One method includes: capturing, using a plurality of video cameras associated with an infrastructure inspection unit, two or more videos of infrastructure; accessing, using one or more processors, image metadata indicating a mesh of connected vertices based on the two or more videos; selecting, using the one or more processors, image data of frames of the two or more videos for inclusion in an output based on the mesh; and outputting, using the one or more processors, a photo-realistic image of the infrastructure comprising the image data selected. Other examples are described and claimed.

Abstract: Modular robotic arms include modular sections with a joint and arm segment. In some examples, the modular sections may connect to each other using common tooling flanges, such as a standard robotic interface for an end effector. In the same or different examples, a robotic arm controller is located in a base or other section for the robotic arm. In the same or different examples, a robotic arm joint includes a differential gear with two motors. Coordinated control of the two differential joint motors facilitates both hinge and rotary motion for the joint.

Abstract: A machine vision-based method and system for locating an object within a scene are provided. The method includes uniformly illuminating a target surface of the object within the scene with light having an intensity within a relatively narrow range of wavelengths such that the light overwhelms the intensity of ambient light within the narrow range to obtain reflected, backscattered illumination. The method also includes sensing brightness of the surface due to a diffuse component of the backscattered illumination to obtain brightness information. Backscattered illumination from the target surface is inspected to obtain geometric information. Rotationally and positionally invariant surface albedo of the object is computed based on the brightness and geometric information. The surface albedo and the geometric information may then be used by a matching algorithm.