Capital Science 2008
An Open-Source Virtual Manufacturing Automation Competition Raj Madhavan, Stephen Balakirsky, Chris Scrapper Intelligent Systems Division National Institute of Standards and Technology (NIST) Presented by: Raj Madhavan, Ph.D.
[email protected] March 29th 2008
Capital Science 2008
Automated Guided Vehicles (AGVs) •
Integral component of today’s manufacturing process – Intra-factory transport – Truck-trailer (un)loading – …
•
Typically used in large manufacturing operations – Large passageways – Little clutter – Large amount of infrastructure
Certain commercial hardware, software and tools are identified in this paper in order to explain our research. Such identification does not imply recommendation or endorsement by the National Institute of Standards and Technology, nor does it imply that the hardware, software, or tools identified are necessarily the best available for the purpose.
Capital Science 2008
Small & Medium Manufacturers
• Present unique challenges for operation in unstructured environment, amongst humans, and with dynamics • Significant cost entry barrier drives elimination of expensive infrastructure
Capital Science 2008
Making Manufactures More Competitive
• • • •
Striving to allow increased automation for small- and mediummanufacturers NIST does not manufacture/develop commercial robots or control systems NIST does design performance metrics to allow current and potential endusers to competitively compare technology and promote innovation NIST brings together AGV end users, manufacturers, and researchers to understand potential “opportunities for improvement”
Capital Science 2008
Three Communities; One Purpose
Industry • • • •
Developers
Researchers
Meet w/ industry to understand high-level needs Distill needs into individual performance metrics and competencies Involve researchers to crack hard problems Demonstrate performance to developers through the use of repeatable artifacts
Capital Science 2008
How To Get Researcher Involvement: Competitions! • NIST has a history in designing and operating Urban Search and Rescue Competitions • Promote real-world, relevant research • Prove performance metrics and standards • Provide researchers an opportunity to showcase their work and compete against peers
NIST wanted to organize a similar competition for manufacturing
Capital Science 2008
Motivation and Benefits •
Industry benefits – Bring increased exposure to their research problems – Provide a free workforce to solve these problems – Create a pool of students that are available for absorption into their companies
•
Student benefits – Lower the entry barrier for robotics research through the use of simulation systems and open source software – Facilitate interdisciplinary research among other robotics & automation fields
•
NIST benefits – Provide a test bed for new standards and performance metrics – Provide ties to both researchers and industry
Capital Science 2008
Virtual Manufacturing Automation Competition (VMAC) •
Advancing Robotic Research through an Open Source High-Fidelity Simulation Framework and Competition – – –
PIs: Stephen Balakirsky, Raj Madhavan, Chris Scrapper Manufacturing Automation Regional Competition One of the 10 proposals that was funded out of 33 submitted
•
Joint NIST, IEEE Robotics and Automation Society effort
•
7 teams selected – Morgan State University, George Washington University, University of Maryland Eastern Shore, George Mason University & Hood College – Faculty: 7 Students: 11
• •
Tutorial and workshop held at NIST in Oct’07 Follow-up workshop held at Hood College in Feb’08
Capital Science 2008
Accepted Teams University (Dept.)
Faculty
Email
GMU (CS)
1. Zoran Duric
1. Eugene Yu
GWU (CS)
2. Rahul Simha
2. Darby Thompson 3. Alex Stefanovski
[email protected]
GWU (MAE)
James D. Lee
Xiaowei Zeng
[email protected]
UMES (EAS)
3. Ali Eydgahi 4. Payam Matin
4. Mark Bivens 5. Chris Epps 6. Lee Long
[email protected]
Hood College (CS)
5. George Dimitoglou 6. Xinlian Liu
7. Brandon Burger 8. Kristopher Reese 9. Eric Walton
[email protected]
Morgan State (ME)
7. Keith Hargrove
10. Therese Andeme 11. Kashun Davis
[email protected]
All participating teams received: • • • •
Student(s)
Linux Machines with USARSim & MOAST A Dell Computer & Monitor Mouse, Keyboard & Joystick Unreal Game Engine
[email protected]
Capital Science 2008
Competition Basics • Teams will be composed of 1 - 3 autonomous unit loader vehicles • Task will be factory logistics – Move packages from a distribution point to several dispersed locations – Multiple paths to destination, some large enough for two-way traffic, others only large enough for a single vehicle – Largest number of correctly delivered packages wins
• Task decomposition – – – –
Path following Docking with roller tables Route planning Traffic management
Capital Science 2008
Software Infrastructure • USARSim – Developed as a NSF funded research tool by CMU and UPitt – Now an open source project managed by NIST – International development community – Physics based simulation – Provides the embodiment and environment for the competition
Capital Science 2008
Scoring Metrics
• Competition consists of “main-event” as well as “best-in-class” competitions • Main-event scoring – Points awarded for successful pick-up and delivery of packages – Points subtracted for dropped packages and collisions, and incorrect deliveries
Capital Science 2008
Best-in-class Path Following • Robot must drive down a series of passageways at specified velocity • Each passageway narrower then the last • Points awarded for distance traveled • Points subtracted for deviations over boundary and in velocity • Test run with and without ground truth
• •
Green lines show commanded path and passageway boundary Red line is actual robot path
Capital Science 2008
Best-in-class Map Building • Technology needed for initial system setup and during normal operation • No ground truth • Scoring leverage RoboCup – Metric quality – Skeleton quality – Multi-vehicle fusion
Capital Science 2008
Best-in-class Docking • Team must dock AGV with loading station – Tested over constellation of starting locations – Free space for maneuvering reduced after each series of runs – Ground truth provided
• Points awarded for successful dock – Time and collisions affect score
Capital Science 2008
Beyond this Competition … •
• • •
Finalists of the competition will demo their results at ICRA 2008 in conjunction with the ICRA Robot Challenge ICRA 2008 Robot Competition Workshop Special session at PerMIS on VMAC competition Competition at ICRA 2009 (Kobe, Japan)
•
NIST – Implementation on real robots in factory settings – Expand range of robots and metrics employed in the competition
Capital Science 2008
Timeline • Final Competition: April 18th 2008 (a month before ICRA’08) • Present Research & Outcomes – Special Session at the Performance Metrics for Intelligent Systems Workshop (Raj Madhavan: Program Chair): August’08 – Follow-up Competition: September’08