System and method for computing grasps for a robotic hand with a palm

A non-transitory computer-readable medium comprising program instructions for automatically grasping an object with a desirable palm grasp, comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape not restricted to a fixed set of shape primitives;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a palm and one or more fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the palm and fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input a description of a specific external wrench that applies to the object;
• computing by the processor a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected grasp configuration;
• comparing by the processor the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choosing by the processor one of the selected grasp configurations with the best quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the palm and fingers of the robotic hand.

A method for automatically grasping an object with a desirable precision grasp, comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape not restricted to a fixed set of shape primitives;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a plurality of fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input a description of a specific external wrench that applies to the object;
• computing by the processor a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected configuration by solving a force optimization problem;
• comparing by the processor the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choosing by the processor one of the selected grasp configurations with the best quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the fingers of the robotic hand.

A method for automatically grasping an object with a desirable precision grasp, comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape not restricted to a fixed set of shape primitives, said object and description each including one or more compliant surfaces;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a plurality of fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input a description of a specific external wrench that applies to the object;
• computing by the processor a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the configuration;
• comparing by the processor the quality measure of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choosing by the processor one of the selected grasp configurations with the best quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the fingers of the robotic hand.

A system for automatically grasping an object with a desirable palm grasp, comprising:

• a robotic hand comprising a palm and one or more fingers;
• a processor configured to:
• receive as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape not restricted to a fixed set of shape primitives;
• automatically select a plurality of grasp configurations of the robotic hand, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the palm and fingers on the object's surface as so described;
• receive as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receive as an input a description of a specific external wrench that applies to the object;
• and
• compute a quality measure for each grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected grasp configuration;
• compare the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choose one of the selected grasp configurations with the best quality measure;
• and
• command the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the palm and fingers of the robotic hand.

A method for automatically grasping an object with a desirable grasp, comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape not restricted to a fixed set of shape primitives;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a plurality of fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input a description of one or more external wrenches that may apply to the object;
• for each grasp configuration, computing by the processor a set of first quality measures, one for each external wrench, representing an ability of the robotic hand to hold the object against each of the external wrenches;
• for each selected grasp configuration, selecting by the processor a second quality measure which is the largest of the set of first quality measures for the selected grasp configuration;
• comparing by the processor the second quality measures of all of the selected grasp configurations to determine which grasp configurations have the best second quality measure;
• choosing by the processor one of the selected grasp configurations with the best second quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the fingers of the robotic hand.

A system for automatically grasping an object with a desirable precision grasp, comprising:

• a robotic hand comprising a plurality of fingers;
• a processor configured to:
• receive as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape not restricted to a fixed set of shape primitives;
• automatically select a plurality of grasp configurations of the robotic hand, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the fingers on the object's surface as so described;
• receive as an input a coefficient of friction at each of the contact points in the selected grasp configuration;
• receive as an input a description of a specific external wrench that applies to the object;
• and
• compute a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected grasp configuration by solving a force optimization problem;
• compare the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choose one of the selected grasp configurations with the best quality measure;
• and
• command the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the fingers of the robotic hand.

A system for automatically grasping an object with a desirable grasp, comprising:

• a robotic hand comprising a plurality of fingers;
• a processor configured to:
• receive as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape not restricted to a fixed set of shape primitives;
• automatically select a plurality of grasp configurations of the robotic hand, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the fingers on the object's surface as so described;
• receive as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receive as an input a description of one or more external wrenches that may apply to the object;
• compute for each selected grasp configuration a set of first quality measures, one for each external wrench, representing an ability of the robotic hand to hold the object against each of the external wrenches;
• select for each grasp configuration a second quality measure which is the largest of the set of first quality measures for the grasp configuration;
• compare the second quality measures of all of the selected grasp configurations to determine which selected grasp configurations have the best second quality measure;
• choose one of the selected grasp configurations with the best second quality measure;
• and
• command the robotic hand to the object using the chosen grasp configuration by making contact between the object and the fingers of the robotic hand.

A method for automatically grasping an object with a desirable palm grasp, comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a palm and one or more fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the palm and fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input a description of a specific external wrench that applies to the object;
• computing by the processor a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected configuration;
• comparing by the processor the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choosing by the processor one of the selected grasp configurations with the best quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the palm and fingers of the robotic hand.

A method for automatically grasping an object with a desirable palm grasp, comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object being of an arbitrary 3D shape, and said description comprising one or more surface patches, point clouds or polygonal meshes;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a palm and one or more fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the palm and fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input as an input a description of a specific external wrench that applies to the object;
• computing by the processor a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected grasp configuration;
• comparing by the processor the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choosing by the processor one of the selected grasp configurations with the best quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the palm and fingers of the robotic hand.

A method for automatically grasping an object with a desirable precision grasp, comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object and description each being of an arbitrary 3D shape;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a plurality of fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input as an input a description of a specific external wrench that applies to the object;
• computing by the processor a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected configuration by solving a force optimization problem;
• comparing by the processor the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choosing by the processor one of the selected grasp configurations with the best quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the fingers of the robotic hand.

A method for automatically grasping an object with a desirable precision grasp comprising the steps of:

• receiving at a processor as an input a description of the surface of an object, said object being of an arbitrary 3D shape, and said description comprising one or more surface patches, point clouds or polygonal meshes;
• automatically selecting by the processor a plurality of grasp configurations of a robotic hand, the robotic hand comprising a plurality of fingers, each grasp configuration specifying a hand configuration and joint variables of the hand, and contact points for the fingers on the object's surface as so described;
• receiving at the processor as an input a coefficient of friction at each of the contact points in the selected grasp configurations;
• receiving at the processor as an input as an input a description of a specific external wrench that applies to the object;
• computing by the processor a quality measure for each selected grasp configuration representing an ability of the robotic hand to hold the object against the specific external wrench in the selected grasp configuration by solving a force optimization problem;
• comparing by the processor the quality measure of each of the selected grasp configurations to determine which selected grasp configurations have the best quality measure;
• choosing by the processor one of the selected grasp configurations with the best quality measure;
• and
• controlling by the processor the robotic hand to grasp the object using the chosen grasp configuration by making contact between the object and the fingers of the robotic hand.