Robot control at singular configurations
AssigneeMassachusetts Institute of Technology
InventorsDavid Verdi, Neville Hogan
Life of this patent
Estimating the term…
Abstract
According to some embodiments, a method includes: receiving an endpoint impedance matrix representing a desired stiffness or damping at the robot endpoint; reflecting the endpoint impedance matrix to an equivalent joint-space matrix associated with one or more of the robot joints, the equivalent joint-space matrix having a nullspace corresponding to near-zero-valued eigenvalues; generating a nullspace-filled impedance matrix from the equivalent joint-space matrix based in part on replacing the near-zero-valued eigenvalues with selected finite positive real values; generating a robot control law using the nullspace-filled impedance matrix; and using the robot control law to control the robot.
Claims19
A computer-implemented method for controlling a robot having an endpoint and a plurality of joints providing more than three degrees of freedom (DoFs), the method comprising:
- receiving an endpoint impedance matrix representing a desired stiffness or damping at the robot endpoint;
- reflecting the endpoint impedance matrix to an equivalent joint-space matrix associated with one or more of the robot joints, the equivalent joint-space matrix having a nullspace corresponding to near-zero-valued eigenvalues;
- generating a nullspace-filled impedance matrix from the equivalent joint-space matrix based in part on replacing the near-zero-valued eigenvalues with selected finite positive real values, wherein the replacing of the near-zero-valued eigenvalues with the selected finite positive real values includes, in response to detecting that the robot is approaching a kinematic singularity, increasing one or more eigenvalues from a threshold value to the corresponding selected finite positive real value;
- generating a robot control law using the nullspace-filled impedance matrix;
- and
- using the robot control law to control the robot.
A computer-implemented method for controlling a robot having an endpoint and a plurality of joints providing more than three degrees of freedom (DoFs), the method comprising:
- selecting impedance parameters for each of a plurality of impedance modules, wherein each of the plurality of impedance modules specifies a robot behavior associated with a task;
- for one or more of the impedance modules:
- receiving an endpoint impedance matrix representing a desired stiffness or damping at the robot endpoint,
- reflecting the endpoint impedance matrix to an equivalent joint-space matrix associated with one or more of the robot joints, the equivalent joint-space matrix having a nullspace corresponding to near-zero-valued eigenvalues,
- generating a nullspace-filled impedance matrix from the equivalent joint-space matrix based in part on replacing the near-zero-valued eigenvalues with selected finite positive real values, wherein the replacing of the near-zero-valued eigenvalues with the selected finite positive real values includes, in response to detecting that the robot is approaching a kinematic singularity, increasing one or more eigenvalues from a threshold value to the corresponding selected finite positive real value, and
- defining an impedance control law for the module using the nullspace-filled stiffness matrix and the impedance parameters selected for the module;
- generating a composite control law for the robot using impedance module control laws defined for the impedance modules;
- and
- using the composite control law to control the robot.
A system comprising:
- a robot having an endpoint and a plurality of joints providing more than three degrees of freedom (DoFs);
- and
- an impedance controller configured to:
- receive an endpoint impedance matrix representing a desired stiffness or damping at the robot endpoint;
- reflect the endpoint impedance matrix to an equivalent joint-space matrix associated with one or more of the robot joints, the equivalent joint-space matrix having a nullspace corresponding to near-zero-valued eigenvalues;
- generate a nullspace-filled impedance matrix from the equivalent joint-space matrix based in part on replacing the near-zero-valued eigenvalues with selected finite positive real values, wherein the replacing of the near-zero-valued eigenvalues with the selected finite positive real values includes, in response to detecting that the robot is approaching a kinematic singularity, increasing one or more eigenvalues from a threshold value to the corresponding selected finite positive real value;
- generate a robot control law using the nullspace-filled impedance matrix;
- and
- use the robot control law to control the robot.
Dependencies are inferred from the claim text.