- Number
- 11701774
- Published
- 2023-07-18
- Filed
- 2020-12-16
- Assignee
- Disney Enterprises, Inc.
- Inventors
- Rinke; Tanner et al.
- CPC
- B25J9/161; B25J9/1669; B25J9/1638; B25J9/1635; B25J9/1605; B25J9/163
- Verdict
- High Hardware
- Source
- Google Patents · FreePatentsOnline
The keeper's note
Learning-based real-time vibration-suppressing robotic control (granted).
Abstract
A robot control method, and associated robot controllers and robots operating with such methods and controllers, providing real-time vibration suppression. The control method involves learning to support real-time, vibration-suppressing control. The method uses state-of-the-art machine learning techniques in conjunction with a differentiable dynamics simulator to yield fast and accurate vibration suppression. Vibration suppression using offline simulation approaches that can be computationally expensive may be used to create training data for the controller, which may be provide by a variety of neural network configurations. In other cases, sensory feedback from sensors onboard the robot being controlled can be used to provide training data to account for wear of the robot's components.
Background
BACKGROUND 1. Field of the Description (1) The present description relates, in general, to robots (or “robotic systems” or “robotic characters”) and control systems and methods for such robots. More particularly, the description relates to a method of generating control signals, in real-time, for a robotic system or robot (and to controllers implementing such a method and robotic systems or robots with such controllers) that provides, via learning, computational vibration suppression during the operations and movements of the robotic system or robot. 2. Relevant Background (2) Audio-animatronic figures and other robotic systems often suffer from unwanted vibrations during their operations. In particular, robotic systems or robots often will experience undesirable vibrations when undergoing fast and dynamic motions such as may be useful in a robotic character to provide expressive animations. For example, a robot may have an arm or leg that they move quickly from one location to a second location to provide a desired movement or move to a new pose, and the robot's arm or leg may vibrate significantly upon stopping at the second location. This can be undesirable when trying to replicate a particular character's movements, when trying to provide human-like motions, and so on. (3) Robotic systems are typically designed to be as stiff as possible, but, unfortunately, the physical system is rarely sufficiently stiff to behave like an idealized mechanical system whose components are
Claims
1. A robotic system configured for real-time vibration-suppressing control, comprising: a plurality of components connected by mechanical joints, wherein a subset of the components is modeled as rigid bodies and a subset is modeled as flexible bodies and wherein a subset or all of the mechanical joints is modeled as flexible joints; memory storing input comprising unfiltered control parameters specifying motion of the components over a time period; a plurality of actuators operable in response to control signals to impart the motion of the components over the time period; and a controller comprising: a processor communicatively linked to the memory; an optimizer provided by the processor running software, wherein the optimizer generates filtered control parameters defining a retargeted motion for the components by adjusting the defined motion to suppress vibrations, wherein the optimizer is trained via machine learning based on a set of training data prior to generating the retargeted motion, and wherein the controller generates the control signals for operating the actuators based on the filtered control parameters, wherein a first portion of the machine learning is performed prior to operations of the controller to generate the control signals and is halted during the operations to generate the control signals, wherein the first portion of the machine learning comprises comparisons of target states defined by a plurality of input animations defining motions for the components via operations of the actuators and states determined by a differentiable simulator, wherein the differentiable simulator performs a dynamic simulation of the robotic system performing the retargeted motions with the components, and wherein the dynamic simulation predicts the vibrations for the robotic system in performing the retargeted motion. ||
8. A robotic system configured for real-time vibration-suppressing control, comprising: a plurality of components; memory storing input comprising unfiltered control parameters specifying motion of the components over a time period; a plurality of motors operable to impart the motion of the components; a plurality of sensors monitoring operations of the robotic system; and a controller comprising: a processor communicatively linked to the memory; an optimizer provided by the processor running software, wherein the optimizer generates, based on the unfiltered control parameters, filtered control parameters defining a retargeted motion for the components by adjusting the defined motion to suppress vibrations of one or more of the components, wherein the optimizer is trained, prior to generating the retargeted motion, via machine learning based on a set of training data, wherein the vibrations that are suppressed consist of low-frequency, large-amplitude vibrations of one or more of the components of the robot, wherein the controller operates the actuators based on the filtered control parameters, and wherein the training data includes sensory feedback from the plurality of sensors. ||
15. A robotic system configured for real-time vibration-suppressing control, comprising: a plurality of components; memory storing input comprising unfiltered control parameters specifying motion of the components over a time period; and a controller comprising: a processor communicatively linked to the memory; an optimizer provided by the processor running software, wherein the optimizer generates filtered control parameters defining a retargeted motion for the components by adjusting the defined motion to suppress vibrations of one or more of the components, wherein the optimizer comprises a neural network that is trained via machine learning based on a set of training data to suppress the vibrations while concurrently minimizing differences between the filtered control parameters and simulated control parameters in the set of training data, and wherein the neural network includes a final node comprising a differentiable simulator, wherein the differentiable simulator performs a dynamic simulation of the robotic system performing the retargeted motions with the plurality of components, and wherein the dynamic simulation predicts vibrations for the robotic system in performing the retargeted motion, wherein a first portion of the machine learning is performed prior to operations of the controller to generate control signals and is halted during the operations to generate the control signals.