DPRG
DPRG List  



[DPRG] 2015 RoboColumbus

Subject: [DPRG] 2015 RoboColumbus
From: David Anderson davida at smu.edu
Date: Sat Dec 6 12:16:20 CST 2014

James,

Now that I've had a chance to sleep on it, I think your proposal makes a 
lot of sense.

It separates the RoboColumbus challenge into a navigation-task and a 
cone-finding-task.  Those really are two very different problem sets.  
It allows the robot builder to concentrate his energies on navigation 
and, when that is accomplished, later on cone finding. That makes a lot 
of sense to me.  It is the way I did it :)

So I've been working out a way to do this without changing the basic 
design of the contest, and without making the actual competition overly 
long.   Here's my thought.  We break the contest in two: navigation and 
cone-finding.  But we still run the contest the same way.  Each robot 
gets 3 runs.

Any robot that scores a 2 or 3  (i.e., navigation alone, no cone 
touching) is automatically part of the RoboColumbus Level 0 navigation 
contest.  A separate First Prize is awarded in that category.   The rest 
of the scoring categories (4 through 9) are in the cone-finding contest, 
with it's own First Prize.  The Grand Prize, if you will.

Since each robot is allowed 3 runs, any builder can decide if he wants 
to use any (or all) of those 3 runs to just do navigation. jBot could, 
for example, disable its camera and cone-finding behaviours, and try to 
set the best time in the navigation-only for one of its three runs.

Since time is the only differentiator for these robots, it will of 
course turn into a race!  (As I suspect level 9 will do also once we can 
all accomplish the main tasks.)  However, we can always make the course 
more difficult as folks get better, make the bonus cone harder to find, 
add a second bonus cone, etc.  Even the size of the error circle can be 
adjusted as our robot navigation gets better.

As I say, it really comes down to how the awards are determined. Seems 
like rewarding successful navigation separated from cone-finding is 
something very worth doing.

Thanks for the input.

regards,
dpa



On 12/04/2014 03:42 PM, David Anderson wrote:
> James,
>
> That's a nice idea.
>
> I think it really comes down to when the awards are handed out. Each 
> possible score is actually a little harder task.  So you can think of 
> the scoring as 7 incrementally harder tasks.  We can award a first 
> place in each scoring category, which I think is a form of what you 
> are proposing.
>
> So a robot without the ability to find and touch a cone, that enters 
> three way point circles, scores 3 points since that is a harder task 
> than entering two.  It would beat a robot that only scored 2, and if 
> those were the only two robots in the competition, it would win!
>
> The scoring hopefully is designed so that robots with the same score 
> have accomplished similarly difficult tasks.  Here's maybe a better 
> graph of that (note: needs mono-spaced font to work):
>
> Points          Target           Bonus         Origin
>              circle  cone    circle  cone    circle cone
> 2              1                               1
> 3              1               1               1
> 4              1       1                       1
> 5              1       1       1               1
> 6              1       1                       1     1
> 7              1       1       1               1     1
> 9              1       1       1      1        1     1
>
>
>
> So as I say, you can think of it as 1 contest with 7 levels of 
> difficulty, or as 7 little mini-contests in which you might be the 
> only contestant in your category.
>
> cheers
> dpa
>
>
> On 12/04/2014 03:03 PM, James LeRoy wrote:
>> I very much like your proposed new competition, but I thought part of
>> the idea was to get entry robots with navigation involved that might not
>> have the cone touching end game ready.  With that type of robot only
>> being able to max out at 3 points it would not be very competitive in
>> the competition.
>>
>> What if it was a two stage event?
>> Stage 1 - navigation to way points only.
>> Stage 2 - navigation and cone touching..
>>
>> You must get two points in stage 1 to compete in stage 2, but an entry
>> robot could still win stage 1 and not compete in stage 2.
>>
>> That's similar to how I thought we would evolve the 6 can competition.
>> Stage 1 - collect cans without an apposing robot
>> Stage 2 - compete head to head
>>
>> -James
>>
>> On 12/4/2014 12:10 AM, David Anderson wrote:
>>> Howdy
>>>
>>> With 6Can competition bearing down upon us, I've already been thinking
>>> about the next outdoor contest. So shift gears with me for a moment and
>>> consider RoboRama 2015s a and b.
>>>
>>> Richard Neveau wrote:
>>>
>>> <snip>
>>> jBot ran a demo of David’s proposal for an alternate contest for the
>>> future by finding the cone and then returning to the starting line. 
>>> jBot
>>> also found the cone after going down the steps (8?) on the deck just 
>>> for
>>> fun.
>>> <snip>
>>>
>>> So .... that is the idea I've been promoting for the next RoboColumbus.
>>> In a nutshell it's this: to start at a location, navigate to a distant
>>> waypoint, and navigate back to the origin and stop. That's what the 
>>> jBot
>>> demo at the contest/gathering at Paul's was trying to show.
>>>
>>> I've described at RBNO and online the advantages of a navigation task
>>> that has to return to home. I'll just add at this point that it also
>>> really fits the name, "RoboColumbus." After all, Magellan died halfway
>>> though his journey. Columbus made it home. ;>)
>>>
>>> So I'd like to formally propose this competition for DPRG 2015 
>>> RoboColumbus:
>>>
>>> -----------------------------------------------------------------------
>>>
>>> DPRG 2015 RoboColumbus Competition.
>>>
>>> 1. Contest: to start at one way point, navigate to a target way point,
>>> and return to the origin waypoint and stop.
>>>
>>> Any robot that completes that task is considered to have completed the
>>> contest. Each run is timed.
>>>
>>> 2. There is a 10-foot radius circle around each waypoint.
>>>
>>> If any part of a robot crosses over that circle, the robot is 
>>> considered
>>> to have arrived at the way point.
>>> Arriving at the way point = 1 point. (first time only)
>>> The robot does not have to stop.
>>>
>>> 3. There is an Orange Traffic Cone at the waypoint at the center of 
>>> each
>>> 10-foot error circle.
>>>
>>> Touching the orange cone and coming to a stop += 2 points. (first 
>>> time only)
>>>
>>> 4. There is a third "bonus" cone located behind some obstacle(s) or in
>>> some hard to reach place.
>>>
>>> Arriving within the bonus cone circle += 1 point (first time only)
>>> Touching the bonus cone += 2 points. (first time only)
>>> Robot can attempt to reach the bonus waypoint either before or after 
>>> the
>>> target waypoint.
>>>
>>> 5. Ties are decided by fastest time.
>>>
>>> -----------------------------------------------------------------------
>>>
>>>
>>> I tried to keep the rules as simple as possible, and also backward
>>> compatible with the RoboColumbus we've already run. I've been mulling
>>> through the math and I think this looks fair to all. It would allow
>>> robots without cameras or other means of locating the cone to be 
>>> able to
>>> run in the contest, if they can hit the error circles. So in that sense
>>> it adds a lower level competition, RoboColumbus --level 0, as well as a
>>> higher level.
>>>
>>> Here are some sample scores:
>>>
>>> 2 points start --> target circle --> start circle and stop
>>> 3 points start --> target circle --> bonus circle --> start circle 
>>> and stop
>>> 4 points start --> target circle --> touch cone --> start circle and 
>>> stop
>>> 5 points start --> target circle --> touch cone --> bonus circle -->
>>> start circle and stop
>>> 6 points start --> target circle --> touch cone --> start circle -->
>>> touch cone and stop
>>> 7 points start --> target_circle --> touch cone --> bonus circle -->
>>> start circle --> touch cone and stop
>>> 9 points start --> target circle --> touch cone --> bonus circle -->
>>> touch cone --> start circle --> touch cone and stop
>>>
>>> The winner is the robot with the most points. Ties go to the faster 
>>> time.
>>>
>>> Talking to Richard and Doug, it seems the hardest part of making a
>>> RoboColumbus contest happen is finding a place to run it. Since I'm
>>> sticking my neck out here anyway, I'll volunteer to try to help track
>>> down some place to hold the next contest.
>>>
>>> happy roboting,
>>> dpa
>>>
>>> _______________________________________________
>>> DPRGlist mailing list
>>> DPRGlist at dprg.org
>>> http://list.dprg.org/mailman/listinfo/dprglist
>> _______________________________________________
>> DPRGlist mailing list
>> DPRGlist at dprg.org
>> http://list.dprg.org/mailman/listinfo/dprglist
>

More information about the DPRG mailing list