Robotic manipulation planning based on probalistic elastoplastic deformation material point method
AssigneeIntel Corporation
InventorsJavier Felip Leon, Ignacio Alvarez, Maria Soledad Elli, David Gonzalez-Aguirre, Javier Turek
Life of this patent
Estimating the term…
Abstract
A robotic manipulation planning system, including at least one processor; and a non-transitory computer-readable storage medium including instructions that, when executed by the at least one processor, cause the at least one processor to: process perception data to detect known, familiar, and unknown objects to generate manipulation candidates; filter manipulation candidates against constraints to reduce the manipulation candidates; and determine quality metrics for the reduced manipulation candidates using a soft-body simulation technique.
Claims23
A robotic manipulation planning system, comprising:
- at least one processor;
- and a non-transitory computer-readable storage medium including instructions that, when executed by the at least one processor, cause the at least one processor to:
- process perception data to detect known, familiar, and unknown objects to generate manipulation candidates;
- filter manipulation candidates against constraints to reduce the manipulation candidates;
- and
- determine quality metrics for the reduced manipulation candidates using a soft-body simulation technique, wherein the determining the quality metrics using the soft-body simulation technique comprises:
- simulating a manipulation action on an object model using a probabilistic material point method (PMPM) based on object material parameters modelled by random variables distributed by a probability density function with sigma points, wherein the sigma points are propagated through the object model to update the probability density function;
- and
- determining a quality metric of the simulated manipulation action based on the updated probability density function, wherein the quality metric is a function of an elastoplastic deformation penalty term accounting for deformation of the object model.
A non-transitory computer-readable storage medium of a robotic manipulation planning system, including instructions that, when executed by the at least one processor, cause the at least one processor to:
- process perception data to detect known, familiar, and unknown objects to generate manipulation candidates;
- filter manipulation candidates against constraints to reduce the manipulation candidates;
- and
- determine quality metrics for the reduced manipulation candidates using a soft-body simulation technique, wherein the determining the quality metrics using the soft-body simulation technique comprises:
- simulating a manipulation action on an object model using a probabilistic material point method (PMPM) based on object material parameters modelled by random variables distributed by a probability density function with sigma points, wherein the sigma points are propagated through the object model to update the probability density function;
- and
- determining a quality metric of the simulated manipulation action based on the updated probability density function, wherein the quality metric is a function of an elastoplastic deformation penalty term accounting for deformation of the object model.
A robotic manipulation planning system, comprising:
- a processing means for processing perception data to detect known, familiar, and unknown objects to generate manipulation candidates;
- a filtering means for filtering manipulation candidates against constraints to reduce the manipulation candidates;
- and
- a determining means for determining quality metrics for the reduced manipulation candidates using a soft-body simulation technique, wherein the determining means determines the quality metrics using the soft-body simulation technique by:
- simulating a manipulation action on an object model using a probabilistic material point method (PMPM) based on object material parameters modelled by random variables distributed by a probability density function with sigma points, wherein the sigma points are propagated through the object model to update the probability density function;
- and
- determining a quality metric of the simulated manipulation action based on the updated probability density function, wherein the quality metric is a function of an elastoplastic deformation penalty term accounting for deformation of the object model.
Dependencies are inferred from the claim text.