No announcement yet.


  • Filter
  • Time
  • Show
Clear All
new posts

  • Regulations/Tasks

    Hello all,

    I was hoping to get a clarification on Requirements/tasks for 2020. We had been in contact before and were told that the requirements would be relatively the same. Can anyone from RoboNation provide information on any changes to regulations or validate that none are changing? If not when will the official regulations/tasks be made available?

    Any help is appreciated,

    John Johnston

  • #2
    Hey JohnJohnston


    Short answer:
    • Nothing will change
    Longer answer:
    • Sizes and weights of the vehicle and the torpedoes/markers will not change.
    • Pinger won't change
    • The location (San Diego - NIWC PAC / TRANSDEC) won't change


    Short answer:
    • Tasks will change...
    • ...but not too much
    • However, I'm always open to suggestions
    Longer answer, There will be:
    • a gate you must pass through
    • buoys you will need to touch
    • bins to drop your two markers into
    • cutouts to shoot your two torpedoes through
    • an "octagon" to surface and manipulation tasks
    • path segments to point toward the beginning tasks
    • two pingers to help locate the end tasks

    With a roaring 20's theme (Prohibition/Bootlegger and G-men), there will be two separate "paths" to take.

    There will be a Bootlegger "path" and a G-man "path"
    At the start, your vehicle will decide if it wishes to complete the tasks as a Bootlegger or as a G-man.
    There will be a way to complete each task as a bootlegger or as a g-man.
    More points are awarded for completing each task based on your initial choice.

    Hopefully that answers your questions. If not please don't (anyone) hesitate to ask more questions.


    • #3
      Suggestion for Daveman :

      I think an interesting task that somewhat bridge the gap between the easier tasks (gate, buoy) and the harder tasks (torpedoes, octagon) are visual inspection tasks.

      As an example, consider if right after the gate there's a small structure with an easily changed state, such as the color or a reversible image. Then you have to behave based on that state. So, maybe you have a reversible arrow which points either right or left, then you hit the buoys in an order corresponding to what you observe.

      some advantages:
      • Easy to implement
      • Not that much harder than buoys, but definitely harder. Represents an attainable next step for teams on the bubble of being really strong, or a differentiating factor for top teams
      • Promotes true autonomy and not ded-reckoning
      • Uses sensors everyone (should) have.
      • Real-world AUV/ROV tasks are often visual inspection


      • #4
        Seth - MuddSub
        Interesting suggestions. I've tried similar things like this before, but I'm always open to more ideas!
        • You haven't mentioned how to implement it, so how do we know if it is easy to implement?
        • I do try to incorporate different point values for the buoys, so this is in line with how I think.
        • Nothing wrong with ded-reckoning, but I do get your point.
        • True
        • also true.


        • #5
          We think the task is interesting and would indeed bridge the gap between the easier tasks and harder tasks. And we also think it would fit the theme with the path to choose between the Bootleger path or the G-man path.

          But if we want to avoid dead reckoning (and our team wants to), we can't rely on a visual indicator that can be seen from the side of the Transdec before the sub start its run. This is the advantage of using a pinger to have a random task and prove autonomy, cause we can't know.

          We are trying to think of a solution to offer on the implementation because we don't think it is easy to create this task using just basic sensors everyone can have and yet make it detectable only for the submarine.


          • #6
            I agree with SONIA. One catch of the course right now is that we mostly know the course before the robot is launched. Our robot this past year was not programmed to pass through the small section of the gate, it was programmed to drive right or left based on our command. Some tasks I would like to see more of is tasks where the state is unknown before the sub is launched.

            Here are some ways to implement this randomness after the robot is launched.
            • Each task could have a light next to it that can be toggled in the way pingers are toggled. The state of this light could determine how the task should be performed. For example: a red light means go through the right side of the gate, a green means go on the left side, no light means go under the gate. Underwater screens displaying different images are also an option but are more difficult to implement and detect.
            • Another way this can be implemented is to have divers change the state of some tasks during the run. This would probably require more divers though and would most likely cause issues.
            • This randomness can also be achieved by having information on the course hidden. This could be as simple as a tall box on the bottom of TRANSDEC which contains a color, letter, or image. Due to the fact it is in a tall box, you can not see the contents unless directly above the box. The robot could then look at this information and once again perform some task differently based on the contents of the box.

            On top of the randomness, one feature of the course this past year I really enjoyed was the style points through the gate. It might be fun to give point bonuses if a sub does a task upside down or in any other weird state.

            Of course all of these options will not be required to get points for any task and are just a way to get bonus points. That is another feature of this past competition I enjoyed, having multiple ways to get points for each task.


            • #7
              Club Etudiant SONIA
              I (always) look forward to your suggestions

              • lights can be an issue underwater, they can get washed out by the sunlight, and/or just be generally hard to see.
                • They require power (either shore power or internal batteries)
                • Depending on the depth, the colors will change.
                • I have some ideas for the future, which involve color, but not lights, but I haven't thoroughly through them through
                • These aren't negative against lights, just things to consider
              • We have a hard enough time getting enough divers in the water to just watch the vehicle(s). With the turnover we get, I try to keep it simple for the divers, and only have them set things up just before the run. We don't want a diver to swim out in front of a vehicle to change the course.
              • Hidden is good, but things only stay hidden for a short while. Something that can change automatically is another story.
              Interesting suggestion for conducting tasks in different orientations (but might be hard to implement).

              I also like to see multiple ways to complete a tasks as well as multiple ways to get points for a task.

              Thanks for your suggestions!


              • #8
                The suggestions to have the AUV demonstrate more autonomy and less dead reckoning are good and represent the point of the whole competition. Keep in mind though that the more complex the setup for the field to achieve a high degree of autonomy will make it more difficult for teams to duplicate this at "home". If that is what the game is then that is what the game is. It is important to push the envelope and try to achieve more difficult things. It is also important to make sure that enough teams can have some degree of success and for spectators to have something more that watching AUVs that barely more or do anything. The trick is to get the right balance. That falls to you Dave...

                I have seen that lights underwater for signaling are poor as you mentioned. Solid panels of a certain color are more reliable, but again are vulnerable to the effect of the position of the sun and the reflections of the sun off the panel. Between the two I would select the panel...I also think that having an AUV follow these color cues should not be mandatory but rather award more points or bonus points. By doing this you will allow more teams to be able to compete at a level that can be challenging, even if you are dead reckoning.

                If teams are more dependent on dead reckoning then another way to challenge them is hold them to a higher standard of accuracy in their navigation....Yes it is not the true autonomy that people are referring too, but dead reckoning across transdek and hitting an object that is only 2 feet wide after traveling over 400 to 500 feet with multiple turns and several 360 degree rotations is a pretty neat trick that many teams would be hard pressed to do. Maybe for these teams make target 1 foot wide? Maybe have them return through the starting gate and then return to the octagon for a second time?

                All in all autonomy is the name of the game, picking the right cues is critical. Teams need to be able to duplicate as much as possible what is going to be done at the competition. I know its never exactly the same, but if teams are to get in the water early and practice, practice, practice, then it needs to be attainable.

                Have fun with this Dave, we await you pronouncement with great anticipation!


                • #9
                  "Promotes true autonomy and not ded-reckoning" Seth - MuddSub
                  "The suggestions to have the AUV demonstrate more autonomy and less dead reckoning are good and represent the point of the whole competition." Ledge

                  I disagree with the premise that a robot that relies on dead-reckoning is not autonomous. Autonomy is a spectrum. As far as I know, an official definition of autonomy for AUVs does not exist. However the Society of Automotive Engineers has formally defined it.


                  Using these definitions, an AUV that has no human intervention and utilizes dead-reckoning would be classified as having Level 4 autonomy (out of 5). Very few (if any) of the RoboSub AUVs would qualify as level 5. It's fine if as a community we want to work towards level 5 autonomy. I believe that was the intention of including aspects of the competition like the coin flip, style points, and the random active pinger. Is it possible to achieve a level 5 system using just dead-reckoning? No, I don't believe so, but it is a valid tool that an AUV can use to help reach that goal. We should continue to promote and reward teams that develop systems with higher levels of autonomy, but I don't believe it is fair to discredit those who successfully utilize valid methods that one might consider a lower-level form of autonomy. While the AUVs at RoboSub vary where they fall in the spectrum of autonomy, all of the AUVs at RoboSub are truly autonomous.


                  • #10
                    dzf209 Yes, perhaps my statement was a bit crude. However, I stand by the sentiment. Regardless of the definitions of autonomy, what I think makes robosub such an interesting problem is the challenge of having the robot make informed, real-time decisions about interacting with the course. Of course, even that will require some amount of ded-reckoning to accomplish. So, the "true autonomy" phrase was wrong, but I still think the more interactions with the course we can promote, the better.

                    That's also why I think it's important to introduce some more medium-difficulty interaction tasks which can be done with minimum hardware but clever software.

                    Also, Daveman , I do realize that this isn't so easy to implement after all. Perhaps there's some way it can be manipulated from the side of the pool with cables or something.