Simulated box placement for algorithm evaluation and refinement

A method to control a robot, comprising:

• storing for each of a plurality of items a set of attribute values representing one or more physical attributes of the item;
• using at least a subset of the set of attribute values to simulate the successive placement of items comprising the plurality of items on a pallet or other receptacle, including by:
• determining, iteratively, for each next item a corresponding placement location at which to place the item on a simulated stack of items on the pallet or other receptacle;
• using the attribute values and a geometric model of where items have been simulated to have been placed to estimate a state of the stack of items after each of at least a subset of simulated placements, wherein estimating the state of stacks of items includes introducing noise in the simulated placements based at least in part on a predefined noise profile;
• using the estimated state to inform a next placement decision, including determining a plan for placing a next item;
• and
• causing a robot to be controlled to place the next item according to the plan;
• wherein the steps of determining for each next item a corresponding placement location and estimating the state of the stack until all of at least a subset of the plurality of items have been simulated as having been placed on the stack.

A computer program product to control a robot, the computer program product being embodied in a non-transitory computer readable medium and comprising computer instructions for:

• storing for each of a plurality of items a set of attribute values representing one or more physical attributes of the item;
• using at least a subset of the set of attribute values to simulate the successive placement of items comprising the plurality of items on a pallet or other receptacle, including by:
• determining, iteratively, for each next item a corresponding placement location at which to place the item on a simulated stack of items on the pallet or other receptacle;
• using the attribute values and a geometric model of where items have been simulated to have been placed to estimate a state of the stack of items after each of at least a subset of simulated placements, wherein estimating the state of stacks of items includes introducing noise in the simulated placements based at least in part on a predefined noise profile;
• using the estimated state to inform a next placement decision, including determining a plan for placing a next item;
• and
• causing a robot to be controlled to place the next item according to the plan;
• wherein the steps of determining for each next item a corresponding placement location and estimating the state of the stack until all of at least a subset of the plurality of items have been simulated as having been placed on the stack.