| NIST Disaster City Event |
Last Modified: 2007-06-28
For this travel journal, we traveled to Disaster City in College Station, Texas to work with the good folks from NIST, FEMA, and several nationally recognized fire-fighters with the goal of creating standards for fielding and evaluating rescue and response robots. Since responders are scattered all over the country and robots that can be fielded are few and often require training to operated, this exercise has been managed by NIST for several years in an effort to create a standard set of tests to validate and qualify different robots for use in response circumstances like natural disasters, terrorism, and hazardous material accidents. In the end, a user or organization like FEMA or DHS can rely on these standardized tests to understand the capabilities of robots entering the market for these uses.
Acroname has been involved in the past 3 such NIST exercises which have taken place in the desert near Las Vegas and the past two were on the TEEX facility near Texas A&M in College Station, Texas. For this event, we worked with a new higher-power Hokuyo laser mounted on a tilt plate that allowed the laser to automatically be tilted. This allows for full 3-D scans of areas in and around the disaster scenarios where the robots were operating.
The idea behind these 3-D laser scans is to provide a rapid model of a space that is completely unknown for responders and robots alike to use in better understanding the dynamics of a building, cave, or structure that is completely unknown to the responders. The scans can be used by structural engineers in these response teams to assess structural integrity, they can be used by responders cutting shoring in the field to secure the structure for entry, and finally, the scans can be used to determine sizes, shapes, and types of structures in an unknown space.... including potential victims.
We put together a couple of sample scans to show both a photo of a room as well as the scanned data for the same. Both scenarios included a small child and an infant dummy to allow us to determine scale as well as to see how difficult it would be to identify people in these scans.
Below is a photo of the scene we scanned. The photo is taken from roughly the same position that the scan was performed from. Since the scan is nearly a complete sphere of scanning, the scanner need only be place around the center of the space to get the best scan.
Sample scenario we set up to perform a scan
After taking photos of this room, we then placed the scanner just around the corner of a doorway looking into the room. This placement was intentional to emulate a pole-mounted scanner that could be fed down a corridor or chamber where the scanner is used to probe around the corner into this room.
This scan can be used with our Symonym simulation environment on most platforms. To view the scan, you will need to download the Free Symonym Player at www.symonym.com . This player is completely self-contained and once it is downloaded and placed onto your local hard-disk, you can add the scan file below in the folder containing other .stick files under the aSymonym sub-directory. Then just run the Symonym application from the aBinary directory and you should be able to select one of the scans you have downloaded to the aSymonym directory.
These scans are pretty big. The number of data points is on the order of 0.3 to 0.5 Mega points. Notice that at first, the scan doesn't make much sense. Once you spin things around a bit (using the arrow keys and shift key), the 3-D data jumps out and starts to make good sense.
Try using the image and the laser scan data side-by-side to learn more about the scene. The first responders were pleased to see that you can "fly up" above the data to see down from a different perspective than that of the original scan. In addition, they could readily see structure and assess where they may need shoring or other support should they need to enter this room or "void".
This scan was done in a pancake collapse with a car and the same two victims from the previous scan. We placed the laser scanner in two locations and did scans to see if the victims could be picked out quickly by the responders. The results varied but it was easiest to pick out the victims with the viewer driving the display of the points. As they spin around the data, they can understand the spacial arrangement of the data. Give these scans a try and see if you can find the victim lying on the door jamb in each scan. Once you see them, they jump out at you.
Satoshi sets up the scanner for the first scan of the car
The laser scanning kept us pretty busy during the 3 days in the warm and humid Texas weather. Concurrently, there were a number of standardized tests running with various mobile robots being teamed with first responders as drivers. Each of these tests was designed to stress a particular part of the robot, the driver, or the interaction between the two.
Manual Dexterity test where the robot must identify objects in peep holes
Most of the robots in the exercises are BEDS (big, expensive, dumb, simple) which proves to work pretty well for most of these tasks being tested. Very few exhibit any form of autonomy so most are strictly teleoperated. Tracked arrangements with various flippers prove to be most common and almost all are outfitted with cameras, some manipulators, and even sensor packages for sensing hazardous materials, etc. None of these robots is cheap, so it was very good of the vendors to lend the machines for testing. Probably all vendors came away with a much better understanding of the weak links in his or her robot platforms as shown by new operators and rigorous testing.
A Talon robot outfitted with a hazardous material sensor deck
While most of these robots borrow heavily from either the design of military tanks or radio controlled monster trucks, there were a couple of truly innovative robots demonstrated that "broke the mold" and offered some new ideas for solving problems of responder robotics.
The first was a quad-rotor "helicopter" UAV which was a mixture of carbon fiber, hand-built DC brushless motors, smart sensors with active control loops for stabilization, and innovative control structure. The two most impressive things about this robot are how quiet it is in operation and, more importantly, the fact that this thing was running nearly continuously for 4 days in hot weather with inexperience pilots, a couple of incidents with neighboring trees, and it continued to run flawlessly, with frequent feedings of Lithium Ion.
The quad-rotor which is under 1 meter in diameter and about a kilogram
The other innovator was Dr. Satoshi Tadokoro who quietly wandered the various scenarios with his optical snake robot. This marvelously simple, yet clever robot uses a bucket of small pager motors along a cable assembly to create a vibrating snake approximately 10 meters long. The outer portion of this assembly is wrapped in a bias-ply fabric that effectively provides "cilia" down the entire length of the snake biased in one direction. The vibration causes the entire assembly to crawl along the ground in a fascinating manner. Near the end are two articulating joints allowing for steering and driving the snake is a simple matter as the video camera at the end provides a simple perspective and the active illumination at this end provide light even in tight, dark quarters.
With nearly no training, it is relatively easy to navigate through cracks and crannies and since the video is wired and not RF-dependent, the image returned is crystal clear.
Dr. Satoshi Tadoroko with his clever snake robot
So... no first responder exercise would be complete without some sort of incident where people respond. This event included a failure of an autonomous system in the form of a Segway scooter which decided that moving along fast up a hill was somehow out of bounds and so the scooter decided to stop.... instantly. Canadian Alex was dumped unceremoniously onto his face on the pavement and was fortunate to come out OK and remarkably remained upbeat about the entire incident.
The strange thing was, we were in Disaster City working with fake victims in all sort of rescue and response scenarios so when you saw a bunch of first responders crowded around a victim lying in the street, you really didn't take a second look.
Alex sports the effects of his recent bludgeoning by robot
This exercise was a coordination of several groups. First, thanks to Dr. Satoshi Tadokoro of Japan who brought the new, yet to be release Hokoyu laser scanner and tilt mechanism. Next, thanks to the TEEX group at Disaster City for putting together both a first class facility as well as herding cats for the days we were there. Special thanks to the first responders who were willing to sit with the "nerds" and learn a bit about upcoming technologies, your input was invaluable. Vendors who are willing to come and put the robots in others hands like hovering parents at a prom are to be congratulated. Finally, thanks to NIST for coordinating, sponsoring, and managing the standards effort in response robotics.
| voice: 720-564-0373, email: email@example.com, address: 4822 Sterling Dr., Boulder CO, 80301-2350, privacy |
© Copyright 1994-2012 Acroname, Inc., Boulder, Colorado. All rights reserved.