Systems and methods for grasping and placing multiple objects with a robotic gripper
A method of grasping and/or placing multiple objects by a gripper of a mobile robot. The multi-grasp method includes determining one or more candidate groups of objects to grasp by the suction-based gripper of the mobile robot, each of the one or more candidate groups of objects including a plurality of objects, determining a grasp quality score for each of the one or more candidate groups of objects, and grasping, by the suction-based gripper of the mobile robot, all objects in a candidate group of objects based, at least in part, on the grasp quality score. The multi-place method includes determining an allowed width associated with the conveyor, selecting a multi-place technique based, at least in part, on the allowed width and a dimension of the multiple grasped objects, and controlling the mobile robot to place the multiple grasped objects on the conveyor based on the selected multi-place technique.
1 . A method of grasping multiple objects by a suction-based gripper of a mobile robot, the method comprising:
determining, by a computing device, one or more candidate groups of objects to grasp by the suction-based gripper of the mobile robot, each of the one or more candidate groups of objects including a plurality of objects;
determining, by the computing device, a grasp quality score for each of the one or more candidate groups of objects;
grasping, by the suction-based gripper of the mobile robot, all objects in a candidate group of objects of the one or more candidate groups of objects based, at least in part, on the grasp quality score determined for the candidate group of objects;
determining, by the computing device, a target object to grasp by the suction-based gripper of the mobile robot, wherein:
each of the one or more candidate groups of objects includes the target object,
a plurality of non-target objects are arranged in a vicinity of the target object, and
determining one or more candidate groups of objects to grasp by the suction-based gripper of the mobile robot comprises:
determining, by the computing device, to exclude a non-target object from a candidate group when the non-target object satisfies at least one criterion; and
including in the one or more candidate groups of objects only non-target objects that are not excluded.
2 . The method of claim 1 , wherein the at least one criterion includes a face of the non-target object being misaligned by a threshold amount from a face of the target object.
3 . The method of claim 2 , further comprising:
defining an alignment window based on the face of the target object; and
determining the non-target object is misaligned by a threshold amount from the face of the target object when at least one corner of the face of the non-target object is outside of the alignment window.
4 . The method of claim 3 , wherein defining the alignment window is further based on the face of the non-target object.
5 . The method of claim 1 , wherein the at least one criterion includes at least one dimension of the non-target object being above a threshold dimension.
6 . The method of claim 1 , wherein the at least one criterion includes at least one dimension of the non-target object being unknown.
7 . The method of claim 1 , wherein the at least one criterion includes the non-target object having a different longest dimension than a longest dimension of the target object.
8 . The method of claim 1 , wherein the at least one criterion includes the non-target object having a dependency on at least one other non-target object determined to be excluded from a candidate group.
9 . The method of claim 1 , wherein determining a grasp quality score for each of the one or more candidate groups of objects comprises:
for each object in the candidate group, using a physical model of object-gripper interaction to evaluate a grasp quality of the object-gripper interaction; and
determining the grasp quality score based on the grasp quality determined for each of the objects in the candidate group.
10 . The method of claim 1 , further comprising:
determining, by the computing device, for each object in the grasped candidate group of objects, a grasp quality; and
releasing one or more objects from the suction-based gripper based, at least in part, on the grasp quality for at least one object being below a threshold grasp quality.
11 . The method of claim 10 , wherein releasing one or more objects from the suction-based gripper comprises releasing each object having a grasp quality below the threshold grasp quality and/or releasing a first object having a grasp quality above the threshold grasp quality and being located adjacent to a second object having a grasp quality below the threshold grasp quality.
12 . The method of claim 1 , further comprising:
determining, by the computing device, a depth of the plurality of objects within each of the one or more candidate groups of objects,
wherein grasping all objects in a candidate group of objects of the one or more candidate groups of objects is further based, at least in part, on the depth determined for each of the one or more candidate groups of objects.
13 . The method of claim 12 , wherein determining the depth of the plurality of objects within each of the one or more candidate groups of objects comprises:
modeling dependencies between objects in a stack of objects as a directed acyclic graph, wherein each node of the directed acyclic graph represents an object in the stack and each directed edge between nodes in the directed acyclic graph represents an amount of physical blocking between the nodes;
determining the depth of each of the plurality of objects in a candidate group of objects based as a longest path between an object with no blocking by other objects in the stack and the node in the directed acyclic graph representing the object; and
determining the depth of the plurality of objects within a candidate group of objects based on a sum of the depths of each of the objects in the candidate group.
14 . The method of claim 1 , further comprising:
determining, by the computing device, an object placement property for at least one object of the plurality of objects within each of the one or more candidate groups of objects,
wherein grasping all objects a candidate group of objects of the one or more candidate groups of objects is further based, at least in part, on the object placement property determined for each of the one or more candidate groups of objects.
15 . The method of claim 1 , wherein the suction-based gripper includes a plurality of suction cups, and the method further comprises:
assigning a seal confidence to each of the plurality of suction cups; and
controlling operation of each of the plurality of suction cups based, at least in part, on the assigned seal confidence for the suction cup.
16 . The method of claim 15 , wherein controlling operation of each of the plurality of suction cups comprises controlling a leak detection process and/or a cup retrying process.
17 . The method of claim 15 , wherein assigning a seal confidence to each of the plurality of suction cups comprises:
assigning a first confidence value to a suction cup when an inner diameter but not an outer diameter of the suction cup is within a face surface of an object to be grasped; and
assigning a second confidence value to the suction cup when both the inner diameter and the outer diameter of the suction cup is within the face surface of the object to be grasped.
18 . A mobile robot, comprising:
a suction-based gripper; and
at least one computing device programmed to:
determine one or more candidate groups of objects to grasp by the suction-based gripper, each of the one or more candidate groups of objects including a plurality of objects;
determine an object placement property for at least one object of the plurality of objects within each of the one or more candidate groups of objects;
determine a grasp quality score for each of the one or more candidate groups of objects; and
grasp, by the suction-based gripper, all objects in a candidate group of objects of the one or more candidate groups of objects based, at least in part, on the grasp quality score determined for each of the one or more candidate groups of objects and the object placement property determined for each of the one or more candidate groups of objects.
19 . The mobile robot of claim 18 , wherein
the suction-based gripper includes a plurality of suction cups, and
the at least one computing device is further programmed to:
assign a seal confidence to each of the plurality of suction cups; and
control operation of each of the plurality of suction cups based, at least in part, on the assigned seal confidence for the suction cup.
20 . A method of grasping multiple objects by a suction-based gripper of a mobile robot, the method comprising:
determining, by a computing device, one or more candidate groups of objects to grasp by the suction-based gripper of the mobile robot, each of the one or more candidate groups of objects including a plurality of objects;
determining, by the computing device, a grasp quality score for each of the one or more candidate groups of objects;
grasping, by the suction-based gripper of the mobile robot, all objects in a candidate group of objects of the one or more candidate groups of objects based, at least in part, on the grasp quality score determined for the candidate group of objects;
determining, by the computing device, for each object in the grasped candidate group of objects, a grasp quality; and
releasing one or more objects from the suction-based gripper based, at least in part, on the grasp quality for at least one object being below a threshold grasp quality, wherein releasing one or more objects from the suction-based gripper comprises releasing each object having a grasp quality below the threshold grasp quality and/or releasing a first object having a grasp quality above the threshold grasp quality and being located adjacent to a second object having a grasp quality below the threshold grasp quality.