Placement-Informed Robotic Picking for Bin Packing Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic pick and place systems often operate in isolation, leading to inefficient space utilization and suboptimal object placement, particularly in tight bin packing scenarios, as they do not effectively integrate pick and place processes.
Innovation Solution
A method that computes object masks from a pick scene and place region masks from a placement area, calculating costs for object-region pairs based on place constraints to select the most suitable object for picking, thereby integrating pick and place processes for efficient object selection and placement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If pick and place systems operate independently with limited interaction, then system complexity is reduced, but space utilization efficiency deteriorates
Solution Approach 1:
The patent merges the pick system and place system into an integrated system where the pick system includes a place predictor that uses place bin state information to influence object selection. This combination allows the systems to share data and coordinate decisions, improving space utilization without proportionally increasing complexity.
Solution Approach 2:
The place predictor performs preliminary analysis of the place bin state before the pick system selects objects. By predicting future place bin states and identifying suitable placement regions in advance, the system optimizes space utilization before the actual picking operation occurs.
2Ease of operation
If pick system selects objects without considering place bin state, then pick operation simplicity is maintained, but placement optimization deteriorates
Solution Approach 1:
The place predictor acts as an intermediary between the pick system and place system. It receives place bin state information and generates predictions that guide object selection, mediating between the simplicity of independent operation and the need for optimized placement without requiring direct complex interaction between pick and place systems.
3Productivity
If traditional bin picking is performed without place information integration, then operational speed is maintained, but packing efficiency deteriorates
Solution Approach 1:
The system performs preliminary prediction of place bin states and identification of suitable placement regions before the actual picking operation. This advance preparation enables optimized packing decisions without adding significant time to the operational cycle, maintaining productivity while improving packing efficiency.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for performing placement informed robotic picking of objects includes acquiring a first image (302) of a pick scene (220) including a number of objects (218) and acquiring a second image (308) of a placement area (224) that receives objects picked from the pick scene (220) by a robot (202). Object masks (306) are computed by performing instance segmentation based on the first image (302). Place region masks (312) are computed by clustering locations in the second image (312) based on a height level from a floor of the placement area (224). A cost is computed for respective object-region pairs (306, 312), each object-region pair (306, 312) defining a pairing between an object mask (306) and a place region mask (312). The cost is defined at least in part by a place constraint. An object (218) is selected to be picked from the pick scene (220) by the robot by selecting an object-region pair (306Sel, 312Sel) based on the computed cost.