Rules & Tasks

[RoboNation]

UPDATED JANUARY 31, 2019

---

CURRENT VERSIONS (RULES & TASKS)

See attached for the latest version of the Rules and Tasks documents for the 2019 RoboSub Competition:

• RoboSub 2019 Task Ideas

---

TEAMS NOTE: See below for previous versions of this document.

[Stephen Cronin]

I wanted to provide some feedback from the perspective of a student who has competed in multiple years of sub and is now advising a team of all freshman in their efforts in the competition. Firstly, I realize these are preliminary ideas and are in no way finalized so please do not take any criticism I have as an attack on the ideas, I realize they have a long way to go.

The first thing that stuck out to me is that it seems on a year over year basis the theme of the competition is beginning to govern core details of the tasks at hand rather than provide a fun medium to present updates to tasks. Each year, the similar tasks come out in concept and at a higher-level logic, but in many respects operate very differently under the hood. What I see mostly from this document is the focus on complex manipulation. This in itself is great but can have wildly different implementations depending on what is being manipulated. I see multiple sliding tasks, but also the grabbing of a vampire (arbitray shape?). This to me would suggest an entirely different kind of gripper than used in years past and may become irrelevant in another years’ time if the theme doesn’t merit that type of manipulation. Overall, this may not be a bad thing, it provides fresh engineering challenges, however the cost (time, knowledge, polish) of reengineering regularly must be considered.

The next point I would like to raise is the direction of difficulty creep in the competition. I understand the goal of increasing difficulty year over year. Technologies are dramatically increasing the capabilities that robots can have at a price point that’s accessible (Jetsons for instance). That being said, it seems like the goal of increased difficulty this iteration is taking an entirely different approach than the competition last year. As I see it, the focus last year was increasing difficulty by increasing the value that big picture autonomy had on the competition. There was value in not just stepping through the linear course in a task by task manner, rather the way to maximize points was to visit multiple places to gather coins in order to deposit more. This raised the difficulty in a manner that allowed teams to rise to the challenge or completely ignore the bigger picture. While you could argue the same for this year, you don’t have call a triangle buoy side, the difficulty is raised in a manner that isn’t as transferable from year to year. This to me is the fundamental question that I would like to see teams and the competition organizers provide input on. What is robosub becoming? Should we be building more capable subs or subs with the ability to understand the implications of their actions in an operating environment?

The final area I think should be addressed is the distribution of complex robotic capabilities asked of a team. The tasks presented seem to rely heavily on the ability to precisely maneuver or to be able to perform complex manipulation. This at its core is fine, however it simply needs to be acknowledged that in many respects the best way to solve some of these problems is just to throw money at sensors. While students getting access to expensive sensors is inherently not a bad problem, it provides them with an excellent idea of how subs run in the real world, we must also accept that it inherently creates a divide in what portion of the competition succeeds. This is not a guaranteed rule certainly, I’ve seen many teams without the funding rise to the challenge and produce capable systems without all the latest and greatest sensors, however it is something that should be considered. Is there a way that teams with less can be at less of a disadvantage?

I don’t want this to come across as too negative, I just want to provide some insight and hopefully start a conversation about what is best for the students and competition moving forward. The competition is a fantastic tool to teach students and develop real-world engineering skills, and I don’t want to overshadow that. I’d be happy to discuss any points I’ve made with anyone who feels differently.


Stephen Cronin

[Daveman]

Stephen Cronin Thanks for taking the time to pass on your knowledge. First and foremost, I never take offense, and always appreciate the critique.

I've always tried to tie the theme into the competition as closely as possible while maintaining interesting tasks. What do you consider complex manipulation? I try to make sure there are multiple way to accomplish every task, and we have had many past tasks where teams had to pick up an object, so a team can use/modify what they have worked on in the past.

There is a difficulty creep, but I try to be conscious of including both "easy" and difficult tasks to encourage new teams to compete while keeping it interesting for "older" teams. There are multiple ways to increase the difficulty ("random" starting direction, passing through the gate, calling the buoy side, choosing a random task, etc.), however it's true that there currently isn't anything that "links" tasks together. I am open to suggestions.

That is a debate that seems to come up every couple of years, teams with lots of funding vs teams with little funding. Throwing money at a sensor is not the magic bullet everyone thinks it is. There is a lot more effort that needs to be put into the sensor(s) to trust what it is saying. There is a reason why so many of these tasks are visual, a lot can be accomplished with a cheap camera. Success typically boils down to time in the water. The more a team has in-water time, the better they (usually) do.

I look forward to a continued dialog.

FYI, we used to put this into the rules of the competition:

Goals of RoboSub
The goals of the AUVSI Foundation’s student competitions are to provide opportunities for students to
experience the challenges of system engineering, to develop skill in accomplishing realistic missions with
autonomous vehicles and to foster relationships between young engineers and the organizations
developing and producing autonomous vehicle technologies.

The primary emphases of the AUVSI Foundation's student competitions are learning and outreach.
These events are not grand challenges designed explicitly to progress the state-of-the-art. The objective
is to produce the people who will push the envelope in the future. Major innovations may be spawned
in these events, but this is a by-product, not an objective. Most important are gaining an appreciation
for the trade offs inherent in any system design and the lessons learned in transitioning from a working
bench prototype to operating reliably in the real world.

When competitiveness and collegiality are in balance, learning is maximized. The AUVSI Foundation's
competitions strive to maintain this balance. The nominal winners are those teams that have scored the
most points. The real winners are all those participants who have learned something lasting about
working together to create an autonomous system that accomplishes a challenging mission in a complex
environment.

The legacy of the student competitions can be found today throughout government and industry.
Employers and venture capitalists seek out prospects with the kind of resourcefulness and team
management experience that former competitors offer.

[Stephen Cronin]

Daveman Let me just go through your questions one by one.

On the subject of complex manipulation, I would agree that most of the tasks can use a simple gripper (something like what bluerobotics sells) and be able to accomplish the task with good motion capabilities. Where I think it moves beyond this capability into what I would consider complex is things like the proposed vampire picking up. What we have right now (assuming I've interpreted the proposed task ideas correctly) is a irregularly shaped vampire stored within a box. Rightfully so this is a more complex manipulation challenge, but solving this one would favor such devices as a soft gripper with the ability to conform a shape or to mitigate impacting the casket. Perhaps I'm wrong here, perhaps there can be a approach which scales the difficultly on the same gripper.

On the subject of task linkage, robosub unfortunately (but realistically) with always suffer from the problem of demonstrating partial competition of linked tasks. RobotX this year for instance enabled showing the judges the output of one task so even if done incorrectly, it was provable that if the linked task was completed correctly based on a bad linkage, performance could be accurately judged. I will try to provide some ideas however. If the goal is to make robots be designed to be cognoscente of the whole world rather than what's in front of them, copies of the initial buoys could be deployed further along the course. Teams can simply go through the gate can earn points, or earn more by exploring the full course to hit buoys in a specific order. Other options would dropping based in a bin based on the vampire was on the buoy. I do agree that vision should be the expected sensor of teams. Let me put some more effort into looking through past years and other competitions for some inspiration.

Finally, I absolutely agree that you are meeting the goals of robosub. I understand the balance that you need to take on every decision, and I can certainly a product of the positive learning environment that sub has fostered. I hope I can provide a perspective to help ensure that in the future.

[Daveman]

Stephen Cronin Wait until you see the vampire before making any calls about difficulty.

It's harder to link tasks and give partial credit since the vehicle can't communicate their intent (except for colored lights, which teams have done) while underwater.

I look forward to reading your ideas and thoughts.