Project Virtual Manufacturing and Automation Challenge

Jacobs University EECS COSYP CASE Legal

 

Summary

Palletizing, or packing rectangular boxes of various sizes onto pallets, is a frequently encountered task in many commercial scenarios, e.g., shipment of goods. An extension of this problem is the automated placement of boxes on pallets, henceforth task completion, by means of industrial robot arms. The palletizing and task completion problems are treated in the context of the IEEE ICRA 2012 Virtual Manufacturing Automation Competition. We approach the palletizing challenge by a winner-takes-all strategy, where multiple heuristics are evaluated against the given datasets. Our results show a performance comparable to that of the commercial software benchmark from the previous competitions.

Result Videos

This video shows results obtained during the VMAC 2012 competition.

Results

Competitors at the Virtual Manufacturing Automation Competition had to develop software that efficiently packs rectangular boxes of various weight and size on a pallet. The problem of not only fitting as many articles per pallet as possible but to also pack them in a way that increases overall stability, is a frequently encountered task in many commercial scenarios.

Jacobs University participated at this competition for the first time and in all three rounds the achieved scores were higher than the one of the competitors, Teams from Drexel University and Georgia Institute of Technology.

The Automation Group developed a heuristics, which is inspired by methods, how humans would address the task. This intuitive solution was implemented by the master student Johannes Schauer and performed very well during the competition, producing nice pallets, i.e., pallets that are efficiently stacked.

Further Results

While preparing, we have used the examples from last year. Please refer to the following table for an evaluation of our approach. Our results while registering for the event are called Jacobs, which were later on improved to Jacobs2.

Data Set GT Drexel Jacobs Jacobs2
D1R1 84.36 67.59 90.43 94.14
D1R2 28.96 32.94 81.96 84.13
D1R3 9.69 26.45 80.99 80.99
D1R4 40.63 44.46 87.81 88.04
D2R1 78.18 39.61 89.87 92.65
D2R2 21.56 - 80.27 80.64
D2R3 4.80 - 75.69 76.15
D2R4 37.88 - 81.61 83.63

People

Publications

  • Girum Demisse, Razvan Mihalyi, Billy Okal, Dev Poudel, Johannes Schauer, and Andreas Nüchter, Mixed Palletizing and Task Completion for Virtual Warehouses. in Virtual Manufacturing Automation (VMAC '12) Workshop at IEEE International Conference Robotics and Automation, ICRA, 2012 [Get Paper (PDF)] [Talk (PDF)].

Software

Please refer to the sisyphus open source project.

Last changed: 2011-04-05