Competitors:
Welcome to the 1996 International Aerial Robotics Competition (IARC). This is the first in a series of E-MAIL messages that you will receive concerning the competition, its rules, and arrangements leading up to the day of the competition.
E-MAIL will be the primary method for transmitting information to the teams and is likewise the preferred method of communication between the teams and the competition organizers (AUVSI) of which I am your principal contact.
The following is a list of the competitors meeting the entry deadline for this year's competition:
As many of you already know, the 1995 International Aerial Robotics Competition was held in July on the campus of the Georgia Institute of Technology in Atlanta, Georgia, USA. That competition was filmed by Scientific American as part of their "Frontiers" series. The special that they filmed aired on U.S. Public Broadcasting stations on January 17 at 8 PM (eastern standard time). This prime-time show is hosted by Alan Alda (who spent the day of filming here in Atlanta last July) If you missed it, watch for reruns this summer (1996).
The 1996 competition is being held at EPCOT which is the science/technology and cultural center at Walt Disney World in Orlando, Florida. Discovery's "World of Wonder" will be one of the production companies televising the event this year.
QUESTION:
>The way I read it, any image processing that occurs after the vehicle lands is not included in the "elapsed time from autonomous takeoff to autonomous landing". Is this correct?
ANSWER:
That was my intent, but it was also my intent that if non-real time image processing was done, it would be rather "immediate" after landing so the judges would have the answer soon after the images were downloaded from the air vehicle rather than having to wait 30 minutes for some Cray in California to crunch on an FTP'd file. Allowing off-board, after-the-fact processing of the image was mainly to allow teams not to have to employ ultra-wide band down links and to allow the use of a hard-copy printer for display of the map to the judges. But this is supposed to be very soon after landing and in most cases within the same hour time slot. When I get all of the entries in, I want to make this interpretation clear. From a logistical standpoint, the judges need to be able to give a score for a team soon after its final run is completed.
QUESTION:
>Can a vehicle take off and land multiple times for a single round, and if so, does only the airborne time count as the flight duration? This could be an effective strategy for reducing the time in the air, which counts as a penalty to the final score.
ANSWER:
No, the run begins with the take off and ends with a landing. These can only be within the starting area. If you land in the starting area, the run ends. If you land outside the starting area, a foul is declared.
QUESTION:
>I notice that the elapsed time measured in seconds from take off until landing is subtracted from a teams score, whereas last year, it was the elapsed time from leaving the starting area to first disk acquisition in seconds _divided_by_ten. This is now on the order of 20 times the penalty as last year, and leads to some interesting results.
>- if a team has a disfunctional vision system, operating aautonomously and intelligently for more than 30 seconds is not a good idea.
>- if locating drums has a probability of success of less than 50% in 50 seconds per drum, it is better not to try finding drums at all, assuming contents is not identified.
>Is this a mis-print (somehow the / 10 got left off) or is this really what is intended?
ANSWER:
No, time is a greater penalty this year. This is because we are trying to simulate a real-world mission (not just an abstract disk retrieval). The DOE is co-sponsoring the Symposium this year and we wanted to make the mission have a DOE flavor (hence the toxic waste theme). The same basic behaviours are still required of the aerial robots, however. In a real survey, it is desirable to be able to do it in minimal time for reasons of efficiency when surveying large areas.
I am sure that there are local maxima for point scores based on scenarios that I never dreamed of. Flying the mission with a disfunctional vision system would not be something that is desirable anyway. So eventhough more points can be scored for short "blind" flights, that strategy shouldn't win against a longer intelligent flight that is providing good vision data.
QUESTION:
>I'm inclined to think the safe thing is to make no assumptions about the clustering of the drums.
ANSWER:
Good assumption.
The drums will be clustered. Clusting is not defined. This is in accordance with the rule that you quoted which says that the individual drum locations will be an unknown. But the reason that I provided a typical clustering picture was to show that the drums would not be scattered (widely separated) around the arena. The distance between drums could be 2 feet or 6 feet, that will be an unknown, but relative to the size of the arena (60 x 120 feet), they will be clustered. The other unknown is the location of the "cluster" within the arena. All you know is that you won't have to skirt the boundaries to work with the cluster once you locate it.
QUESTION:
>What is the ratio of plastic to steel drums and 25 to 55 gallons drums?
ANSWER:
Unknown by the teams.
QUESTION:
>Are all the Biohazards drums plastic?
ANSWER:
Unknown by the teams.
QUESTION:
>How many samples are there?
ANSWER:
Only one.
QUESTION:
>Will the sample be resting on a steel drum?
ANSWER:
No.
QUESTION:
>Will the labels be flat relative to ground?
ANSWER:
No. The angle of the drum and the curvature of the drum will distort the label for those on drums that are not sitting up-right. Labels will be on the most horizontal protion of the drum, however. Drums won't be at 45 degrees, but will be closer to vertical or horizontal.
QUESTION:
>What is the maximum number of team members?
ANSWER:
No limit.
QUESTION:
>How can award tuition money be used and how is it given out?
ANSWER:
Given as a check usually to the School's department, but the AUVSI tries to do this in a way that prevents the department from reallocating to uses that do not DIRECTLY benefit the aerial robotics team members. The money is intended to be used as an offset to tuition or to improve the vehicle for future competitions. It is not intended for frivolous purposes.
QUESTION:
>A couple of our members have made the case that the final graphical display we are to present the judges could be nothing more that a video snapshot from high above the field, carefully overlaid onto a grid pattern. If the resolution is good enough, you'll be able to make out the labels, thereby fulfilling the requirements. Although I doubt this is your intention, I have been persuaded to get an official answer.
ANSWER:
That is a very good point. This is definitely not the intent. I am looking for machine-generated coordinates such that the system is determining the coordinates, not the judges. Also the machine should determine the identification of the drum labels, not the judges. The object is to create a machine that does the mapping and ID without human intervention. If human interpretation or reduction of the data is required, then that is not autonomous from beginning to end (though obviously in the "end" there must be a human interface for the judges).
QUESTION:
>What exactly will the labels look like. I know they will be negative images of what has been provided in the rules. I know they will be the same size as in the rules. But when we took the negative of those images it became unclear as to whether the white border will be included exclusively, or if there will be some black area surrounding the white border.
ANSWER:
The border will be 8.5 by 11 inches (just like the sheet of paper upon which the rules are printed) and the black rectangle will be white and included within the 8.5 by 11 inch label.
QUESTION:
>Should our lift-off succeed but fail before landing, can we re-start from the failure point using the already gathered data and continue gathering data?
ANSWER:
All restarts must be from the starting area in the corner of the arena. You bring up an interesting point however, and that is what if you locate one of the drums on an initial flight-- do you have to "re-locate" that drum on your next try, or can you remember the location information from your earlier flight? We could simply add the time of all the flights taken to acquire the information incrementally, but the spirit of the mission is to gather the data as if you were really in the field at a toxic waste site, and if your aerial robot failed (i.e., landed in the middle of the site) you might not have the luxury of retrieving it.
So no team is able to benefit from the efforts of another earlier team, the location and orientation of the drums is to be different for each team (some one could "listen in" on your telemetry and get a head start on knowing the configuration of the site were we not to do this). We will shuffle the drums after each flight attempt as well (to assure that no a priori information gathered during aborted runs is of use.
Assume therefore, that the orientation and location of the drums will change randomly between runs. A record of the actual locations will be kept by the judges for comparison with the data that a team gathers on any particular run.
QUESTION:
>Where will the disk be located on a drum? Can it be affixed on the round part of the drum?
Yes, it will be on a drum. It will be on a relatively horizontal portion of the drum (most likely an upright end) so it won't slide off.
QUESTON:
>3) What is the diameter of each drum top/bottom?
ANSWER:
Don't know the answer to this yet. The place that we are getting the drums from is donating them, so we will get what they offer. They will be of the 50 - 55 gallon variety in either plastic (polyethelyene?) or steel. There WILL be some plastic drums, not all will be steel, so be prepared for both types.
QUESTON:
>4) How high can we fly above the playing feild?
ANSWER:
For safety reasons, keep your vehicle under 30 feet. Vehicles perceived to be going higher than 30 feet may have their run terminated by the judges for fear that loss of control might send the craft into people or property. Other restrictions may be placed upon us by Disney that are as of yet undetermined, but any restriction will generally be imposed in the interest of safety.
QUESTION:
>5) What is the planerity of playing field, is it level, is it sand, etc?
ANSWER:
Grass surface. One of the two sites under consideration is extremely level, while the other has a gentle undulation and is slightly crowned in the center. Disney is sending blue prints of the two sites and topography/surroundings for the one chosen will be transmitted to the teams.
QUESTION:
>Is the disk guaranteed to be oriented on top of a drum? (i.e. A flat surface parallel to the ground?)
ANSWER:
Not necessarily on the flat end of a drum. It will be on a relatively horizontal portion of the drum (though most likely an upright end) so it won't slide off.
QUESTON:
>When the rules say that the labels will be visible from above, does this mean from directly above, or might the label be oriented on the side of a drum that might make it only 50% visible from directly above?
ANSWER:
No, it will be facing straight up (being pasted either on an end or side of the drum). Remember that it might be at an angle due to the angle of the drum (drun's "pitch axis"), and it may be distorted by the curvature of the side of the drum. But it won't be on the side of a drum that is facing away from verticle in its "roll axis".
QUESTION:
>Would it be possible for a drum to be totally submerged, so that only say 5 - 10% of the drum is visible?
ANSWER:
No, at least 50 percent of a partially burried drum will be visible. In some cases 100 percent may be visible.
QUESTION:
>What is the color of the drums?
ANSWER:
Black or deep blue.
QUESTION:
>Is the background fresh green grass (same as last year)?
ANSWER:
Yes
QUESTION:
>The radioactive and biohazard labels are being placed on the top of the drums? Or might they be stuck to any position on the surface of the drums ?
ANSWER:
They will be on the upper side of the drums, but the drums may be on their side or at an angle. In any case, the label will be on the most vertically visible portion of the drum.
QUESTION:
>In Figure 3 Example of Typical Drum Arrangement, the stickers can be viewed rather from the side than from the top, contrary to what is stated in the text "visible from directly above".
ANSWER:
Yes, the Figure you are referring to was meant to show a typical arrangement of the drum (standing up, on the side, or at an angle) but was not meant to indicate the label position. Follow the guidance given in the rules WORDING rather than the Figure when considering the label positions. The Figure indicates examples of drum orientation only.
QUESTION:
>Inregard to page 6 of 16 "Scoring", 5.: How accurate is "accurate mapping"? You are specifying within one meter, but relative to what?
ANSWER:
Relative to the actual (True) position of the drum. If a drum is located at exactly x equals 3 and y equals 5, then you need to be able to identify the position to within a one meter circle of that point (so an answer of x equals 3.25 and y equals 4.75 would be acceptable).
QUESTION:
>A 55 gallon drum has a dimension of 2 x 3 ft, so do we map the center of gravity, the position of the sticker(s), just any point within the drum's location or the outline of each drum?
ANSWER:
Good question. Map based on the centroid of the visible (above-ground) shape. that means that a drum burried at a 45 degree angle would have a centroid that is very much different (relative to the drum's center of gravity) than a drum lying half buried on its side (in which case the center of gravity would correspond to the centroid of the exposed portion of the drum as viewed from above).
QUESTION:
>7) Are there any written compendiums detailing the past contest entries, including methodology and how it failed or succeeded? A Web Site? Any personal notes?
ANSWER:
The only compendium of past entries is the AUVSI proceedings for the past two years. Here, papers describe the entries and the intended operation. The actual results are chronicled in various films, but no detailed analysis of what went wrong with different entries has been compiled. You will find that most of the previous teams are more than willing to discuss their entries with you (both prior entries and their present ones as well!).
Georgia Tech Research Institute Aerospace Sciences Laboratory (AERO-CCRF) 7220 Richardson Road Smyrna, Georgia 30080 robert.michelson@gtri.gatech.edu Voice: (770) 528-7568