You are here

29 posts / 0 new
Last post
Posted by:
Bumble Bee
Bumble Bee's picture
Fri, 03/31/2017 - 12:49pm

Random Pinger

Wanted to create a separate thread for all to discuss cos we think this was one of the major shortcomings for last year. We think the random pinger from last year need to be improved as it was boiled down eventually to whether you won in the coin toss. Either of two ways:

1. Reduce random pinger to just 500 points

2. Make it fair and enforce that to get the random pinger 2k you need to do both the pinger and the actual task at the location(torpedo or bins and the pick up task)

 

Posted by:
David Novick
David Novick's picture
Sun, 04/02/2017 - 11:13pm

ping...

Thanks for creating this thread.

  1. I believe if a vehicle completes a random task, that should be worth more that 500 pts.  But I'm open to suggestions.
  2. The rules from last year stated that a vehicle must acquire points in the random task in order to get the "random" points.  I intend on keeping that for this year.  Was there something more that you're asking for?
Posted by:
Cuauv Leader
Cuauv Leader's picture
Tue, 04/11/2017 - 4:02am

Two Active Pingers?

1. Could it be possible to have both pingers active at the same time on different frequencies? To keep the random pinger points you could have a certain frequency as the first target. Getting the human switching pingers out of the loop allows for the subs to potentially make more intelligent decisions on its own since we wouldn't have to know exactly what the sub is tring to do.

2. I can see how Bumble Bee views the random pinger as luck of the draw to some degree. A possible solution would be to give a small bonus (500 pts) for getting points at the random pinger and then a larger bonus (2000 pts) if they switch pingers and get points at the other pinger. This keeps the high reward for being able to do random tasks while limiting the luck aspect if getting points at one of the pingers is harder than the other.

Posted by:
BeaverAUV
BeaverAUV's picture
Fri, 04/14/2017 - 10:24am

Point multiplier?

I agree with much of what's been said. I think the two active pingers is a clever idea, but with up to four courses running, there would be a lot of pingers going off creating a lot of noise, which has caused issues for teams in the past.

One way to deal with the issues Bumble Bee brought up would be to switch the random pinger bonus from an immediate, one-time bonus to a multiplier for all the points you gather by that pinger. So if you chose to use the active pinger enabled you'd get something like double(?) all the points you earn at that pinger. This would deal with the 'Luck of the draw' issue, and provide a pretty substantial award for going for the random pinger. It also forces more interaction at each task to get the points.

Posted by:
David Novick
David Novick's picture
Sat, 04/15/2017 - 1:54pm

...ping...

During semi-finals, we have 8 pingers in the water.  I'm sure you don't want different frequency pingers to ping at the same time.  I'm definitly sure you don't want the same frequency pingers to ping at the same time.  In order to get all 8 pingers ringing and deconflicted, there would be a large delay between pings for the pinger your listening to.  Unless I could change the frequency of each pinger (potentially in the works), teams would learn which is the first task and potentially dead recon to it.

CUAV:  What's the different between specifying a frequency and having a human switch the pinger as far as intelligent UAV decisions?

BB: Do you believe the random pinger boiled down to a coin toss because it was "easier" to get points within one task than another to get the random bonus points?

 

Posted by:
Cuauv Leader
Cuauv Leader's picture
Wed, 04/26/2017 - 12:37am

Intelligent AUVs

Ah, didn't realize it wasn't currently possible for the pinger frequencies to be switched on the fly. 

Having the AUV make intelligent planning decisions on its own is difficult enough when the world around it is more or less static outside of its own actions. Introducing more variability in the environment makes this task more difficult. The AUV has to either assume or detect (not trivial if both pingers are identical frequency) that the pingers are switched when it wants to travel to the second pinger. This shouldn't cause issues if all goes according to plan, but limits the ability for the AUV to make an audible on the fly and abort a task early etc...

Posted by:
AUV IITB
AUV IITB's picture
Wed, 04/19/2017 - 6:40am

Pinger

1.We would like to point out here that the random pinger points are associated with the submarine's capability of localising a sound source correctly and not manipulating the environment, and to remove the issue (coin toss) raised by BumbleBee a possible solution would be to grant a small bonus for getting at the random pinger and then a larger bonus if after switching other pinger too is localized (this would limit the luck aspect if getting at the second pinger is harder as pointed out by Cornell). By adding a condition of completing the associated tasks there is no further improvement as once a pinger localised (by fluke or by true autonomy) doing the tasks will not prove how the vehicle got there.
2.There could be some other measure to prove this,

a)perhaps like when the vehicle is moving towards a particular pinger (say it is in within a certain radius), the judges could or could not switch the pinger, depending on which the vehicle will or will not have to change its path. The second switch will then be as before requested by the team.
b) A third waypoint pinger may be added (slightly offset from the path between both the pingers) which may or may not be switched on depending on judges discretion which will decide whether the AUV's path from pinger A to B is straight or via the third pinger.

But in both these alternatives the extra time some AUVs take depending on whether the judges make the extra switch or not will have to be somehow compensated.

3.Apart from this, having multiple pingers (8) run at the same time may definitely add noise, and hence it would be better to have a human switch. Also, I too have the same question for CUAUV as posted by Dave.  

 

Posted by:
David Novick
David Novick's picture
Sat, 04/22/2017 - 5:13pm

Random Number Generator

I think I'm finally tracking what everyone is saying about a coin toss...That even if all a team is going to do is the octagon, and they choose to do a random pinger, they have a 50/50 shot at getting the "correct" pinger turned on for their task and potentially getting the "random" points (a little slow on the uptake on this one).

I'm going to introduce a new rule for the random pinger that states that a vehicle must be able to acquire points in each of the pinger sections (i.e. drop markers for one, surface for another), in order to be able to choose the random pinger points.

I'll give it more thought on how to make this even less of a coin toss and more of a reward.

Posted by:
Boulder Robosub
Boulder Robosub's picture
Sat, 04/22/2017 - 9:37pm

Just to dig into the intent

Just to dig into the intent of the rule, our understanding was that the random pinger points was an atempt to discourage deadreconing the octagon? is there perhaps a better way to discourage deadreconing and still reward localizing the pinger? 

At the same time we think that rewarding teams that go to the effort of being able to determine which tasks to complete without being told in advance is valuable. 

Posted by:
David Novick
David Novick's picture
Sat, 04/29/2017 - 2:07pm

Intent

To me, deadreconing is a valid form of navigation.  So not to discourage deadreconing, but to award teams that can sense and act on the pinger.  I'm always looking for ways to award autonomous behavior.

CUAS:  I understand.  If all the pingers were on, then the vehicle could autonomous decide, "I'm done here, on to the next task,"  instead of having to wait for a silly human to make the switch (and then sense the switch).  True, not a trivial task, but one that can be done, I believe.

 

How about something like this:

500 points for securing points at the first random task.

1500 points for securing points at the second random task.

Posted by:
Bumble Bee
Bumble Bee's picture
Mon, 05/01/2017 - 8:13am

Have been fighting with the

Have been fighting with the new robonation forums(had a bug in the browser which needed the deletion of cache). Sorry for our delayed responses. We agree with the option : I'm going to introduce a new rule for the random pinger that states that a vehicle must be able to acquire points in each of the pinger sections (i.e. drop markers for one, surface for another), in order to be able to choose the random pinger points.

Was gonna suggest that to make it fair for both options of the pick up and torp/bins task and we see value in the random pinger enabling greater autonomy. Dave, what do you mean by first random task and second random task? For each of the task we would need to prolly define what does it mean when you complete a task for random pinger. Our suggestion is for torp/bins you either fire a torp in any hole/drop any marker in any bin and for the pick up task to just pick up and surface with the object. From this level of task completion, the AUVs would have alrdy demonstrated the level of required autonomy. 

On the 8 pingers active problem, we didn't face major issues and was able to localise all the time. 

 

 

Posted by:
David Novick
David Novick's picture
Fri, 05/05/2017 - 5:32pm

Random Pinger

Sorry to hear you were having trouble with the forum.  Glad you where able to get it sorted out.

I believe we are pretty much saying the same thing.  There are two pinger tasks: Cultiave Pearls/Battle Squid (Torpedoes/Bins) and Classify/Rescue (Octagon).  When they are randomized, the first random task woud be the one with the active pinger (could be Torpedoes/Bins or Octagon).

Similar to last year's rules, in order to get points for the random pinger task, you would have to get points at each task.  So, for example, if the random selection chose the octagon as the first task, that would be the active pinger.  The vehicle would have to do something to acquire points in that area to get the 500 points for a random task (surface in the octagon, pick up samples, put them down, etc.).  Once points have been secured, the team captain could request a pinger change, and the vehicle would move to the second random task (in this case, the Torpedoes/Bins).  The vehicle would have to do something  to acquire points in this area to obtain the 1500 points for a random task (fire a torpedo through a cutout, drop a marker in the bin, etc.).

 

Hopefully, that all makes sense.

Posted by:
Bumble Bee
Bumble Bee's picture
Fri, 06/16/2017 - 8:43am

Just to revisit and confirm

Just to revisit and confirm the information that you stated here:  The vehicle would have to do something to acquire points in that area to get the 500 points for a random task (surface in the octagon, pick up samples, put them down, etc.)

Does surfacing only in octagon constitute the 500 bonus points? And what is the minimum requirement? At least pick up samples? 

What we're getting at is that surfacing in octagon is not the same as firing torpedo or dropping a ball in a bin.

Posted by:
Bumble Bee
Bumble Bee's picture
Fri, 06/16/2017 - 8:43am

Just to revisit and confirm

Just to revisit and confirm the information that you stated here:  The vehicle would have to do something to acquire points in that area to get the 500 points for a random task (surface in the octagon, pick up samples, put them down, etc.)

Does surfacing only in octagon constitute the 500 bonus points? And what is the minimum requirement? At least pick up samples? 

What we're getting at is that surfacing in octagon is not the same as firing torpedo or dropping a ball in a bin.

Posted by:
Bumble Bee
Bumble Bee's picture
Sun, 05/07/2017 - 12:17pm

Sounds good! To clarify, does

Sounds good! To clarify, does it mean that the second random pinger task would always be the other variant? That means there won't be a case of two pick up task random pinger or two torp/bin task random pinger.

Posted by:
David Novick
David Novick's picture
Sun, 05/07/2017 - 11:02pm

Ah, good question!

Yes, the second random pinger task will always be the other task  You will never get the same task again.  The randomization is only for the first task.

So if the first random task is the octagon, the second will be the bins/torpedoes.  If the first random task is the bins/torpedoes the second will be the octagon.

Posted by:
David Novick
David Novick's picture
Fri, 06/16/2017 - 6:28pm

Post...???

It said there was a new post from Bumble Bee, but I don't see it.  If I'm missing something, let me know.

Posted by:
Alan Albritton
Alan Albritton's picture
Sat, 06/17/2017 - 12:08am

Bumble bee post

If you scroll up you'll find bumble bees new post. Its a reply under your post titled random pinger in this thread.

Posted by:
Team Kyutech
Team Kyutech's picture
Sat, 06/17/2017 - 5:33am

Pinger Frequencies

Hi! We would like to confirm how the frequencies of the pinger will be decided. Since in the rules, it was specified that the frequencies can be user selected. With that, we have the following questions:

  • Who will decide what frequency will be used, the team or the staff?
  • Will the frequency be decided on the run itself? Meaning, on each run, it's possible that frequencies are different?
  • Will the frequency of the two pingers in one competition area be the same?

We're still kind of confused with the setup. Any help is appreciated!

Posted by:
David Novick
David Novick's picture
Sat, 06/17/2017 - 1:01pm

Pinger Freq

This will be described in the final rules, which really really really (really) should have been out before now (really).

The four quadrants (A, B, C, D) will each have their own frequency and a time at which they ping.  So the pinger for the bins/torpedoes and the Octagon (etc.) will both use the same frequency.  Only one will be on at a time.  I have a switch that can either turn on the bins/torpedo or the octagon What I did last year seemed to work (unless someone speaks up), so I'll use that again.  This is how it works:

  • A: 30kHz, 0 sec
  • B: 40kHz, 1 sec
  • C: 25kHz, 0.5 sec
  • D: 35kHz, 1.4 sec

So the pinger in section A will ping.  0.5 sec later, pinger C will ping.  0.5 sec later (@ 1sec from "start"), pinger B will ping.  Finally 0.4 sec later (@ 1.4sec from "start"), pinger D will ping. 

0.6 sec later (at the top of the next second), pinger A will ping.  The pinger control boxes are GPS synced.

Hopefully that makes sense.  If you have any questions, please feel free to ask.

Posted by:
Team Kyutech
Team Kyutech's picture
Thu, 06/22/2017 - 7:35am

Thank you! Addtnl question: Acoustic Output

Hi Dave! Yes! It makes sense! Thank you!

We have an additional question. We checked the pinger's specs and would like to know what the Acoustic Output (162/168/174/177dB) would be. 

Thanks again!

Posted by:
David Novick
David Novick's picture
Thu, 06/22/2017 - 9:47pm

Power

I typically set it to the maximum (177dB) unless it becomes an issue (and it usually doesn't)

Posted by:
Team Kyutech
Team Kyutech's picture
Tue, 06/27/2017 - 10:16pm

Got it! Thanks!

Got it! Thanks!

Posted by:
Bumble Bee
Bumble Bee's picture
Sat, 06/17/2017 - 12:18pm

Yes dave, we did!! Refer to

Yes dave, we did!! Refer to reply #14!

Posted by:
David Novick
David Novick's picture
Sat, 06/17/2017 - 1:37pm

RNG pinger & Points.

Thank Alan and BB.  It wasn't showing up yesterday, and now it's there twice...

If it's a random task then any points aquired during the task will net 500 points for the first task and 1500 for the second task.  So surfacing in the octagon, surfacing with an object, removing the cover off the bin, dropping a marker in the bin (or border), firing a torpedo through the cutout, etc. will all net the random points if it's a random run.

Are you concerned that surfacing in an octagon is easier than firing a torpedo?  In order to surface in a octagon, you must be able use hear the pinger (hydrophone selection, custom conditioning electronics), determine it's bearing (more custom electronics and custom software), stop when directly overhead and surface (using either the pinger or visual cues to stay on target).  To fire a torpedo (or drop a marker), you must see the object (COTS camera and OS software), align with the target (custom software), and fire/drop (custom electronics).

To me, they are similar, but I'm always open to feedback.

Posted by:
Bumble Bee
Bumble Bee's picture
Mon, 06/19/2017 - 11:13am

Yes that's right! At both

Yes that's right! At both random pinger to torp/bins or pick up, we would have to equally be able to localise to the pinger but it is much harder to fire a torp or drop a marker than to localise with acoustics to a pinger only accurately. Some would even dead reckon there and secure the points for the octagon easily by just surfacing. If you had to do an actual task at either this would make both sides equally difficult. Our suggestion is to make it fair and for the octagon to require at least picking up an object for the bonus to be counted.

 

Posted by:
David Novick
David Novick's picture
Tue, 06/20/2017 - 9:07am

Localize on a pinger

I see a lot of teams struggle to hear the pinger (and not a reflection), localize on the pinger, and then surface above it, so I don't believe it's that easy (correct me if I'm wrong).  Yes, you can dead reckon, but that also has it's issues (knowing you're vehicle well, sensor drift, etc.), and dead reckoning is a valid form of navigation (and a bit of a high risk since your run is over if you surface outside the octogon).  Remember, you need to be able to accomplish tasks in both pinger areas to get random points, so a team that can only dead reckon, can not request a random pinger. 

I also believe that localizing near a pinger (i.e. the bins/torp task) is easier than stopping directly above (again, correct me if I'm wrong).  Some could argue that firing a torpedo is easier, since it involves all COTS/OS software solutions.

For now, I think I'll keep it as-is.  However, if this does become an issue in the future, then I'm happy to revist any alternative solution to help even out the tasks.  As always, I appreciate your input.

Posted by:
Bumble Bee
Bumble Bee's picture
Sun, 06/25/2017 - 4:24am

okay noted dave! Will there

okay noted dave! Will there be points for normal pinger to the octagon?

Posted by:
David Novick
David Novick's picture
Wed, 06/28/2017 - 11:06pm

Points

Not sure exactly the question you're asking.  There are points for surfacing the octagon.  If that's not your question, please feel free to ask again.

Log in to post comments