Defence Business: Team Michigan wins the MAGIC 2010 Challenge | ADM Feb 2011
Gregor Ferguson | Brisbane
Teams from the US swept the board at the MAGIC 2010 (Multi-Autonomous Ground-robotics International Challenge) event held in Adelaide shortly before the LWC.
The results were announced at the start of the conference, with Team Michigan, led by Professor Ed Olson, taking the US$750,000 prize and Team U-Penn from the University of Pennsylvania second and Team RASR (Reconnaissance & Autonomy for Small Robots) third.
The Turkish Team Cappadocia placed fourth while the sole Australian team to make it through to the finals placed fifth.
In all, 23 teams entered the MAGIC 2010 challenge and the five finalists undertook a searching program of tests in Adelaide before the winner was selected.
The Challenge was organised jointly by DSTO’s Land Systems Division and the US Army’s Tank Automotive Research Development and Engineering Center (TARDEC).
“This was a watershed event which will change forever ground robotics,” according to Dr Grace Bocheneck, Director of TARDEC, who described the entries as “outstanding, innovative, challenging, ground-breaking and eye-opening.”
The Challenge was designed to help fast-rack the development of autonomous robotics and the necessary control systems and algorithms.
Currently a single human operator may be able to control two or possibly three robots, at most.
Autonomy would allow teams of robots to act independently tackling the dull, dirty and dangerous jobs which present either an unacceptable risk to humans, or the inefficient use of human resources.
There are currently about 5,000 military robots serving with the US forces in Afghanistan, said the US Assistant secretary of the Army for Acquisition, Logistics and technology Dr Malcolm O’Neill.
“I want this stuff tapped for the US Army and Australia Army,” he declared, adding technology is only beginning to tackle the autonomy issues in robotics.
“We are today where the Wright Brothers were in the early 20th century.”
Major General John Caligari, the Australian Army’s head of Modernisation and Strategic Planning, provided a user’s perspective, commenting that the Challenge had demonstrated “things about the technology we’d never have thought of if we hadn’t seen them for ourselves.”
The winning team Michigan impressed the judges with their system’s reconnaissance and environmental mapping capabilities.
The team consisted of 14 robots with just two operators monitoring their activities.
The robots themselves were simple tracked, skid-steer devices with four 24v electric motors and an Inertial Measurement Unit (IMU) containing a four-axis gyro.
Team leader Ed Olson told ADM the guts of each robot was a laser radar/rangefinder bore sighted to a video camera and an 802.11B wireless protocol with 900MHz radio links and a 115Kbaud data rate so the robot could share telemetry and video.
This is a very bandwidth-efficient network architecture, he told ADM, adding each robot was controlled via an onboard IBM ThinkPad running a Linux-based operating system.
The laser sensor used shape recognition software verified by the video image recognition system.
The robots themselves used a simple graphical image as a unique identifier; the image recognition software was smart enough that operators wore t-shirts with a similar graphic printed on them which instructed the robots to “Follow Me”, thus allowing a single operator to lead a file of robots from one location to another under their own power and without direct human intervention.
Why did Team Michigan win?
Olsen believes it was because they were able to demonstrate high levels of autonomy using more robots than anybody else, while keeping bandwidth requirements low.
The team also adopted an aggressive approach, he said, to building terrain maps using shared robot data; finally, the user interface was simple and effective.
Best of all, the robots were relatively cheap: US$12,000 each, of which around half was accounted for by the laser radar.