Abstract: Systems, devices, and methods for training and operating (semi-)autonomous robots to complete multiple different work objectives are described. A robot control system stores a library of reusable work primitives each corresponding to a respective basic sub-task or sub-action that the robot is operative to autonomously perform. A work objective is analyzed to determine a sequence (i.e., a combination and/or permutation) of reusable work primitives that, when executed by the robot, will complete the work objective. The robot executes the sequence of reusable work primitives to complete the work objective. The reusable work primitives may include one or more reusable grasp primitives that enable(s) a robot's end effector to grasp objects. Simulated instances of real physical robots may be trained in simulated environments to develop control instructions that, once uploaded to the real physical robots, enable such real physical robots to autonomously perform reusable work primitives.

Abstract: Robot control systems, methods, control modules and computer program products that leverage one or more large language model(s) (LLMs) in order to achieve at least some degree of autonomy are described. Robot control parameters and/or instructions may advantageously be specified in natural language (NL) and communicated with the LLM via an NL prompt or query. An NL response from the LLM may then be converted into robot control parameters and/or instructions. In this way, an LLM may be leveraged by the robot control system to enhance the autonomy of various operations and/or functions, including without limitation task planning, motion planning, human interaction, and/or reasoning about the environment.

Abstract: In an implementation, a hydraulic valve includes a valve body, a cylindrical chamber within the valve body, and a cylindrical spool within the chamber. The spool is rotatable between at least a first position and a second position about a longitudinal axis parallel to a longitudinal axis of the chamber. The spool includes a landing and an eccentric channel that extends around a portion of an outer circumferential surface of the landing. In the first position, an inlet and an outlet of the hydraulic valve are fluidly coupled by the eccentric channel to allow a flow of a hydraulic fluid circumferentially around the landing from the inlet to the outlet. In the second position, the inlet and/or the outlet are obstructed by another portion of the outer circumferential surface of the landing to prevent the flow of the hydraulic fluid from the inlet to the outlet.

Abstract: A fluidic tactile sensor includes a core having an outer core portion, an inner core portion, and a first channel having a first opening at a first surface portion of the outer core portion. An elastic skin is disposed over the first surface portion. A cell is formed between the first surface portion and the elastic skin and fluidly is fluidly connected to the first channel. The cell contains a fluid. A contact force applied to the elastic skin produces a measurable change in fluid pressure inside the cell.

Abstract: A flexible joint includes a plurality of links, a plurality of flexures, and at least one cable which can be tensioned or relaxed to cause a bending about a plurality of axes. The links may include a base link; a last link; and a plurality of intermediate links, coupled together by the flexures such that the plurality of intermediate links, the first link, and the last link form a chain of links having the base link at a first end of the chain and the last link at a second end of the chain and each of the plurality of intermediate links is coupled to two other ones of the plurality of links by two flexures. The cable(s) extend through cable pass-through-holes in the links. The links may comprise disks with sloped faces, and may be rotational offset from another around a longitudinal axis. The flexures may comprise living hinges.

Abstract: Human-machine interfaces may capture interactions by humans with robots (e.g., robots with a humanoid appearance), the interactions taking a variety of forms (e.g., audio, visual), and may determine an intent of the humans or meaning of human responses via analysis of the interactions. Intent can be determined based on analysis of aural response, including meaning or semantics and/or tone. Intent can be determined based on analysis of visually detectable responses, including head motions, facial gestures, hand or arm gestures, eye gestures. Responses may be compared for consistency. Humans may be queried to confirm determined intended response.

Abstract: A detachable mechanical eye cartridge includes a plurality of links, a plurality of couplers, a plurality of electronic connection points, and at least one mechanical eye that can be rotatably actuated in at least two rotational degrees of freedom. The detachable mechanical eye cartridge may include actuators for actuating the eye, eyelid structures, an actuator for actuating the eyelid structures, and imaging sensors within the mechanical eye. The detachable mechanical eye cartridge may include one or two mechanical eyes and related structures and may include a set of magnetic couplers that couple to magnets on a robotic head frame to which the detachable mechanical eye cartridge may couple. The detachable mechanical eye cartridge may also have a set of mechanically coupled linkages for controlling the pitch of each mechanical eye with minimal backlash.

Abstract: A robot is provided having a kinematic chain comprising a plurality of joints and links, including a root joint connected to a robot pedestal, and at least one end effector. A plurality of actuators are fixedly mounted on the robot pedestal. A plurality of tendons is connected to a corresponding plurality of actuation points on the kinematic chain and to actuators in the plurality of actuators, arranged to translate actuator position and force to actuation points for tendon-driven joints on the kinematic chain with losses in precision due to variability of tendons in the plurality of tendons. A controller operates the kinematic chain to perform a task. The controller is configured to generate actuator command data in dependence on the actuator states and image data in a manner that compensates for the losses in precision in the tendon-driven mechanisms.

Abstract: Systems, devices, and methods for training and operating (semi-)autonomous robots to complete multiple different work objectives are described. A robot control system stores a library of reusable work primitives each corresponding to a respective basic sub-task or sub-action that the robot is operative to autonomously perform. A work objective is analyzed to determine a sequence (i.e., a combination and/or permutation) of reusable work primitives that, when executed by the robot, will complete the work objective. The robot executes the sequence of reusable work primitives to complete the work objective. The reusable work primitives may include one or more reusable grasp primitives that enable(s) a robot's end effector to grasp objects. Simulated instances of real physical robots may be trained in simulated environments to develop control instructions that, once uploaded to the real physical robots, enable such real physical robots to autonomously perform reusable work primitives.

Abstract: Robots, robot systems, and methods for operating the same based on environment models including haptic data are described. An environment model which includes representations of objects in an environment is accessed, and a robot system is controlled based on the environment model. The environment model incudes haptic data, which provides more effective control of the robot. The environment model is populated based on visual profiles, haptic profiles, and/or other data profiles for objects or features retrieved from respective databases. Identification of objects or features can be based on cross-referencing between visual and haptic profiles, to populate the environment model with data not directly collected by a robot which is populating the model, or data not directly collected from the actual objects or features in the environment.

Abstract: Systems, devices, and methods for training and operating (semi-)autonomous robots to complete multiple different work objectives are described. A robot control system stores a library of reusable work primitives each corresponding to a respective basic sub-task or sub-action that the robot is operative to autonomously perform. A work objective is analyzed to determine a sequence (i.e., a combination and/or permutation) of reusable work primitives that, when executed by the robot, will complete the work objective. The robot executes the sequence of reusable work primitives to complete the work objective. The reusable work primitives may include one or more reusable grasp primitives that enable(s) a robot's end effector to grasp objects. Simulated instances of real physical robots may be trained in simulated environments to develop control instructions that, once uploaded to the real physical robots, enable such real physical robots to autonomously perform reusable work primitives.

Abstract: Provided is a robot that includes: a first sensor having a first output and configured to sense state of a robot or an environment of the robot; a first hardware machine-learning accelerator coupled to the first output of the first sensor and configured to transform information sensed by the first sensor into a first latent-space representation; a second sensor having a second output and configured to sense state of the robot or the environment of the robot; a second hardware machine-learning accelerator configured to transform information sensed by the second sensor into a second latent-space representation; and a processor configured to control the robot based on both the first latent-space representation and the second latent-space representation.

Abstract: Mechanisms or apparatus convert a number of inputs via a number of input members into a number of output movements of an output structure, providing control in three degrees-of-freedom (DOF), for example control over roll, pitch and yaw of the output structure. Inputs may be rotations about a common axis of rotation, for example via a first ring, a second ring, and one or more plates, concentrically array. Rotation of the first ring may control a first DOF, rotation of the first ring may control a second DOF, and rotation of the plate may control all three DOF. Three concentrically arrayed tubular shafts may be employed, providing a through-passage or cable fluid conduit run to accommodate wires, optical fibers, fluid carrying conduits. Such may be particularly advantageous when employed as part of a robot, or other device with a tool or sensor or transducer located at or proximate a distal end thereof.

Abstract: A software compensated robotic system makes use of recurrent neural networks and image processing to control operation and/or movement of an end effector. Images are used to compensate for variations in the response of the robotic system to command signals. This compensation allows for the use of components having lower reproducibility, precision and/or accuracy that would otherwise be practical.

Abstract: A mechanical hand mimics a human hand having similar degrees of freedom and sensory abilities while appearing visually similar to human hand. The mechanical hand comprises a mechanical hand skeleton and resilient elastomer (e.g., silicone) skin that fully encloses the mechanical hand skeleton. The mechanical hand skeleton may advantageously be molded directly into the resilient elastomer (e.g., silicone) skin such that the hand appears, moves, and feels very similar to a real human hand. The mechanical hand may have applications in robotics, for example as an end-of-arm tool or end effector, or may have other applications. Robotic applications may include prosthetics applications.

Abstract: Provided is a process that includes obtaining data indicative of state of a dynamic mechanical system and an environment of the dynamic mechanical system, the data comprising a plurality of channels of data from a plurality of different sensors including a plurality of cameras and other sensors indicative of state of actuators of the dynamic mechanical system; forming a training set from the obtained data by segmenting the data by time and grouping segments from the different channels by time to form units of training data that span different channels among the plurality of channels; training a metric learning model to encode inputs corresponding to the plurality of channels as vectors in an embedding space with self-supervised learning based on the training set; and using the trained metric learning model to control the dynamic mechanical system or another dynamic mechanical system.

Abstract: Provided is a system that includes: a plurality of touch sensors sharing a signal medium, each touch sensor in the plurality being configured to output set of frequencies on the signal medium responsive to being touched, each touch sensor in the plurality being configured to output a different set of frequencies; an analog to digital converter electrically coupled to the signal medium and configured to receive the sets of frequencies from the touch sensors and convert the sets of frequencies to digital representations of the sets of frequencies in the time domain; a processor communicatively coupled to the analog to digital converter and configured to execute a fast Fourier transform of the digital representations from the time domain into digital representations in the frequency domain; and an address decoder operative to transform the digital representations in the frequency domain into identifiers of touch sensors among the plurality of touch sensors.

Abstract: A software compensated robotic system makes use of recurrent neural networks and image processing to control operation and/or movement of an end effector. Images are used to compensate for variations in the response of the robotic system to command signals. This compensation allows for the use of components having lower reproducibility, precision and/or accuracy that would otherwise be practical.

Abstract: A humanoid robot may include one or two eye resembling features in the form of image sensor(s) that capture images of at least a portion of an environment. The fields-of-view of the image sensor(s) will change as a pose (i.e., position, orientation, position and orientation) of the head, torso, or even entire humanoid robot changes. Virtual or augmented representations of the environment may be presented to a human pilot of the robot, where disorienting feeling due to latency in communications can be alleviated by reducing demand on communications channels.