Proposal: Simulation League

[Simon Kohaut]

Dear friends of robot sailing,

I want to propose to you the idea of a simualtion league as an addition to the WRSC. As a mean to test and visualize the software that controls our robot, simulations are an important part of our development process. They enable us to to see the effects of different approaches for planning and control without us depending on the weather and access to water or the need to risk expensive hardware. As simulations are already part of similiar competitions like the RoboCup, I believe it is also a great opportunity for the autonomous sailing community to grow and spark more international interest by implementing and offering such an additional league. It would lower the entry hurdle of autonomous sailing by removing the need to build your own boat and makes it easier to compete on a global scale.

Some of the things it needs to realize a simulation league:

• A dynamics model of a sailboat
• A model or recorded data for the wind
  
• Software that enables custom controllers to read the state of the simulation and send control signals (I believe ROS could be a great framework for this, as it is used widespread)
  
• Challenges/tasks and their respective performance measurement
  
• Ideally some recorded AIS data to have realistically behaving obstacles to avoid
• Visualization

I will be glad to hear your thoughts and ideas about this.

Best regards

Simon Kohaut

Sailing Team Darmstadt e.V.

[Colin Sauze [cos]]

Simon,

I like the idea of doing this and I think it would help teams without access to boats and even help those with boats to develop better control strategies.

I think we need to run it at a separate time to WRSC. As we all know getting a boat working at the competition is hard and adding getting simulated boats working just increases that pressure for teams trying to do both. It also adds more work for the organisers. As it should be easier to do a simulated boat, maybe we could run this a few months before WRSC or even during the winter when a lot of teams can't sail.

Louis Taylor, a former AberSailbot member developed the sailsd simulator (https://github.com/sails-simulator/sailsd) a few years ago and this would make a reasonable basis for this. It has a graphical interface, its quite easy to extract data from it and is designed to be easily extensible. However it's physics model needs some improvement and somebody would need to implement a ROS binding for it. I think there are several other simulators out there and maybe some combination of parts from each of them would be best.

I recorded a few days worth of AIS data during WRSC2018, its available from https://github.com/colinsauze/ASVTrafficSim/blob/master/Southampton-AIS-2018-08-29.txt. I have some code for parsing AIS data and linking it up to a simulator although its not what I would call production ready.

I think the key questions are:

1. Do we want to try/are we able to do this in time for WRSC2019 (or ideally a few months before WRSC2019)?

2. Who is going to be responsible for building the simulator? I'm willing to put in some time towards this, but can't offer nearly enough to complete it on my own.

Colin.

--
You received this message because you are subscribed to the Google Groups "World Robotic Sailing Championships Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wrsc-discuss...@googlegroups.com.
To post to this group, send email to wrsc-d...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/wrsc-discuss/6cf5a897-60bd-4ede-9ca7-603f2a3ec83b%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

-- 
Dr. Colin Sauze
Research Software Engineer
Super Computing Wales Project
Room 2.16, Physical Sciences Building
Aberystwyth University,
Penglais, Aberystwyth
Ceredigion, UK, SY23 3DB

Tel: +44 (0)1970 622774
Webpage: http://users.aber.ac.uk/cos

--------------------------------------------------------------------
Prifysgol Aberystwyth www.aber.ac.uk
Prifysgol y Flwyddyn ar gyfer Ansawdd Dysgu - The Times & The Sunday Times 2019.

Aberystwyth University www.aber.ac.uk
University of the Year for Teaching Quality - The Times & The Sunday Times 2019.

[Yu Cao 曹宇]

Hi Simon,

I like the idea of having a simulation league as part of WRSC. This challenge could help teams to get start with robotics sailing when they don't have access to the hardware. For the teams who have their own boat they can also use the simulator to continue improving the performance during the winter. I also agree  Colin's opinion on the time of the competition, if we are trying to compete annually it is better to not fill all challenge at the same time with WRSC. I would suggest a quarterly competition or a rolling competition (continuous submission and judging) for the simulation challenge. The prize can be given to the team with best overall performance at WRSC competition week every year. 

Beside simulation environment you have already mentioned there is also a Gazebo based simulator [^1]. It provides a visualisation environment that we are current missing. It is great if we can agree on the conventional on the ROS message, so the the simulation backend and visualisation integration and  documentation would be easier. If necessary I could help to setup an repository under WRSC organisation on GitHub for the development. 

We have collected some weather data (wind speed, direction) during this year's competition. It is possible to offer those data as part of competition for the virtual challenge. Four challenges in WRSC 2018 still seems a good start point for the virtual challenge. The same scoring software [^2] may even able to mark virtual attempts with some modification. I am personally interested in implement some Open AI gym compatible APIs for the simulator. 

1: https://github.com/disaster-robotics-proalertas/usv_sim_lsa 

2: https://github.com/WRSC/scoring

Best regards,

Yu Cao 

PhD research student / University of Southampton 

在 2018年10月21日星期日 UTC+1下午7:35:17，Simon Kohaut写道：

[John]

this is a very interesting idea and something I will look more into but currently I am concentrating more on the hardware end of this. Although I have been programming industrial machines with various software for over 40 years I have never done anything in Linux and from the small amount of research I have done it looks like ROS is Linux based. As mentioned I'm working on the hardware, I currently have a basic control systems using a Raspberry Pi and an Arduino. The Pi is a Linux based operating system and I see that ROS can be run on that so I'll look into that more. But a simulated competition would be pretty cool and a great way to learn more.

I am a member of a model yacht club so I understand the boat end of things, but can always learn more. I also hope to be getting a MaxiMOOP hull later this year.

I was wondering though is there a autonomous control forum available? I'm looking for information on programming strategies, sensors and that kind of help as well as some place to ask questions.

Thanks

John  

[Colin Sauze [cos]]

John,

Its not specifically for autonomous control, but you the Microtransat group is a good place to ask these sort of questions. Its got a lot more members than this one (26 vs 122).

Colin

--

You received this message because you are subscribed to the Google Groups "World Robotic Sailing Championships Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to wrsc-discuss...@googlegroups.com.
To post to this group, send email to wrsc-d...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/wrsc-discuss/db226ef5-2828-4537-8122-ace83c3480f0%40googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

[John]

Thanks Colin

I do follow the Microtransat group, very interesting emails. I do find though that most of the equipment and controls are, for lack of a better term, industrial. They're building boats to go across the ocean, I'm sailing around a pond. So most of the equipment is beefier, again for lack of a good term, and the controllers are more satellite based. Not that there is anything wrong with that it's just a different application than I am doing so I haven't been asking.

I will post my questions as I do see that there is a wealth of knowledge there.

Thanks again

John 

On Sunday, October 21, 2018 at 2:35:17 PM UTC-4, Simon Kohaut wrote:

[Sophia M]

Hi All,

Simon, thank you for introducing the simulated sailing idea! The ideas brought together here are very interesting :) Some thoughts from me:

Regarding the timing, I don't think a continuous / several times per year competition is a good idea, at least not for now. I would expect this to need a significant amount of continuous maintenance, and I believe a shorter period will also make it easier to give teams a feeling that they are competing. I agree with the idea of running a simulation competition in Winter, when many teams are not able to go on the water - it would also help with recruiting and familiarizing new team members with sailing robots. Maybe we can start in 2019 with an informal competition, but I would plan for December2019/January/February 2020. This time slot also has the advantage that teams have time to write papers about their simulation and submit to IRSC :)

It sounds like ROS/gazebo would be the best approach, even if teams don't implement their entire work in ROS, they should easily manage to output a ROS message to interface with the competition side. In previous simulator based competitions, OSRF has used docker containers, which also makes it very easy for teams to get started. From personal conversation it sounds like they are willing to share their experience and advice to get started.

The biggest question that I still see open is what form of challenge we want to run. I would not combine AIS data and a full sailing robot simulation immediately, but would instead approach the problem in multiple steps, e.g.:

A) Wind variation during triangle race.

All teams use the same simplified model, they have to make tack decisions etc. to route their boat around a triangle course with changing wind conditions. The ROS output they generate is a rudder and a sail angle. As a final, all boat paths can be shown at the same time. As training data, some wind patterns are given, a secret wind pattern is used as the final.

B) Simulation verification. Control signals of a (simple) given boat, some boat data, and an example recording of wind direction and boat path on a lake are the start point for the teams. Ideally the simulated boat coincides with the class we suggest as a boat for beginners. For the scoring a previously secret boat path and set of rudder/sail setpoints is used. Here not the fastest, but the most realistic boat wins.

C) Obstacle avoidance. The most successful previously achieved boat model, combined with real AIS and wind data. All teams get the same boat, the teams can set a goal heading, aiming to pass through the AIS data without getting too close/colliding with other boats.

This can be escalated further, adding in simulated vehicle component failures, additional boat designs, wave simulation, wind shadow and wakes of boats on AIS, etc.

For the first simulation event we would 'only' need a docker container with ROS, a gazebo simulation of a simplified sail boat, some buoys and a wind pattern - as well a scoring system.

I would suggest splitting the discussion, talking about everything sailing robot specific here, and figuring out the competition implementation on discourse.ros.org (not sure where there exactly), since this will be interesting for other people running simulation based competitions.

Looking forwards to hear your thoughts on this,

Sophia

PS: John, you might get some answers, or pointers towards helpful tutorials, in this group by opening a discussion specifically on your question(s) - e.g. i am fairly sure the Southampton team will have some good advice to get you started with ROS on Linux on a raspberry pi. Generally, people will try to stay 'on topic', so everyone responding to the topic Simon opened here will focus on the simulation league in their responses.