INEOS Team GB

[JonRowe]

With lidar you're getting point measurement across the field, so you don't need spectacular resolution, from a display perspective you could easily use this to set up a 3D model (I don't think you'd "view" it in a traditional sense, you'd look at a visualisation of the data)... I'd say you could also build some interesting "follow the dot" style "this is what the sail is doing, this is what the sail should be doing from vpp" type displays, but the rules forbid that this time around right? Or do they not apply to test boats?

I also breifly wondered if they could use lidar to monitor the seastate around them to feed back into the models... not sure how well lidar works against liquids though

 

[enigmatically2]

However, looking at the examples on the website, it doesn't look like the resolution is spectacular.

LIDAR is pretty accurate (the one you highlight has an angular accuracy 0.01 degrees and range accuracy of 0.1cm. In this context it is more accurate than that because the relative accuracy of adjacent points is more accurate than absolute accuracy.

True you do get some fuzziness when you re-image that back into 3rd, but plenty of good data there. I wonder how much processing they then have on that data (e.g. does it tell you depth/camber, max camber position at various points up the sail). I would expect so. A little bit of short term averaging (over a few seconds) and you will get some pretty accurate answers I would think. And live info. Wonder whether we will see that used live on the race boats themselves

 

[captainSackSparrow]

If it's not banned, I suspect they may be using the LiDAR for wind sensing.

It is used in offshore wind, where getting an understanding of the wind sheer is critical.

Maybe they could use to validate models by collecting flow around sails.

 

[dogwatch]

If it's not banned, I suspect they may be using the LiDAR for wind sensing.

Not allowed, I believe. Protocol 40.2, though it is implicit. Anyone think otherwise?

 

[Stingray~]

If it's not banned, I suspect they may be using the LiDAR for wind sensing.

It is used in offshore wind, where getting an understanding of the wind sheer is critical.

Maybe they could use to validate models by collecting flow around sails.

Was my first thought too. Airflow.

 

[enigmatically2]

Not allowed, I believe. Protocol 40.2, though it is implicit. Anyone think otherwise?

No, looks prohibited to me

They have in any case said the LIDAR is for looking at sails

 

[Stingray~]

No, looks prohibited to me

They have in any case said the LIDAR is for looking at sails

My guess is they are using for a detailed look at airflow, but will be allowed to use it only during testing - never when racing.

 

[enigmatically2]

My guess is they are using for a detailed look at airflow, but will be allowed to use it only during testing - never when racing.

I doubt it if it is the model stated

 

[Stingray~]

I doubt it if it is the model stated

? Missed where the model was stated by Ineos? Perhaps a look at their tech partners might suggest the specialties being used.

Some Boeings are equipped with LIDAR for detecting wind but obviously that's calibrated for detecting at a longer range than what may be being detected here. In some DJI forums the subject of photogrammetry and map making and such gets pretty deep and at the top-end there are methods that rely on (very expensive) drones equipped with LIDAR to create stunningly detailed 3D maps of everything. That is another example of the use of LIDAR sensors, the sensors and applications do vary.

 

[Scillyjosh]

@Scillyjosh might be able to give a few more insights to what these can be used for.

Kind of interesting that INEOS only have one near the starboard jib slot. Whereas Magic have them mounted both sides but further back.

Interesting how cagey he was about how this data was being used. I would have assumed it's piped back to the UK and that's it. But could the trimmers be using lidar 3d images rather than camera feeds? Takes out issues you get from camera's using wide FOV.

In terms of data the lidar takes out a couple of the steps that camera image based systems use. These take distortion out of the image from lens and sensor, then use the trim strips to plot curves and interpolate the sections between to give a 3d shape. Lidar the data that comes out is 3d already.

However, looking at the examples on the website, it doesn't look like the resolution is spectacular.

With the lidar looking up at the rig you would be able to map the sail surface as a point cloud. The segmentation between the sails, hull, spars etc would be challenging. However this could be simplified by having an estimated region the sail occupied, taken from simulation, then searching around that to find the sail shape.

Regarding resolution with those ouster lidars it's not sub millimetre resolution you'll be resolving on a single point but they scan at 10Hz and will be capturing lots of data on the whole surface. Some curve fitting and filtering work can then be done to improve the overall estimate of the shape. This can be augmented in real time with other information i.e. images of trim stripes taken from calibrated time synced cameras.

Looking at what you do with all the data being processed is just as interesting! In other systems using similar sensors (think autonomous cars, drones, robotics etc) a large amount of processing is done locally (edge processing) where mapping algorithms are run. Similarly the sail shapes can be built in near real time and compared to historical test and simulation data. This can also be logged to feed into a simulation later. This could be looking at either how a sail responds to the loads i.e. does it deform as expected or looking at other boat parameters (speed, rig loads etc)

To speed up the simulation loop you can stream all of this to the cloud and run much more computationally expensive processes on it such as running a full simulator or multiple simulators. With this you can tweak simulation parameters to better match the real world or in real time test strategies the simulation predicts will give better performance. Having multiple simulators (testing different physics models) running in parallel allows even more efficiency in simulator tuning.

This would allow two scenarios: one where the simulation is used to train sailors on how to trim the sails. Whilst another strategy would have simulation and sailor working together to test new sailing and simulation strategies.

 

[Scillyjosh]

? Missed where the model was stated by Ineos? Perhaps a look at their tech partners might suggest the specialties being used.

Some Boeings are equipped with LIDAR for detecting wind but obviously that's calibrated for detecting at a longer range than what may be being detected here. In some DJI forums the subject of photogrammetry and map making and such gets pretty deep and at the top-end there are methods that rely on (very expensive) drones equipped with LIDAR to create stunningly detailed 3D maps of everything. That is another example of the use of LIDAR sensors, the sensors and applications do vary.

I use the same ouster lidars as the teams are using at work and they can't see wind, they also see through water so not much info to be resolved there sadly!

The stunningly detailed lidar maps your talking about also tend to look very pretty because they have the photogrammetry overlay which tricks you into thinking the scan is very accurate.

 

[Scillyjosh]

LIDAR is pretty accurate (the one you highlight has an angular accuracy 0.01 degrees and range accuracy of 0.1cm. In this context it is more accurate than that because the relative accuracy of adjacent points is more accurate than absolute accuracy.

True you do get some fuzziness when you re-image that back into 3rd, but plenty of good data there. I wonder how much processing they then have on that data (e.g. does it tell you depth/camber, max camber position at various points up the sail). I would expect so. A little bit of short term averaging (over a few seconds) and you will get some pretty accurate answers I would think. And live info. Wonder whether we will see that used live on the race boats themselves

Those angular resolutions are not correct.
The ouster lidar data output is a 32,64 or 128 pixels vertically and 2048 pixels horizontally.
With a field of view of 45 or 90 degrees you'll get the following resolution vertically:

Whilst all have the same horizontal resolution of:

For range accuracy the ouster OS0 (90° field of view) has the following performance:

Whilst the OS1 (45° field of view) has the following performance:

All data from ouster direct:

https://data.ouster.io/downloads/datasheets/datasheet-rev7-v3p0-os1.pdf

https://data.ouster.io/downloads/datasheets/datasheet-rev7-v3p0-os0.pdf

Total resolution hasn't changed massively between REV01 and REV07 hardware and you couldn't visually identify one model from the other. Most of the changes made are about improving data quality in rain, fog etc.

 

[Stingray~]

Awesome, thanks for the knowledgeable additions and insights into this LIDAR subject.

 

[enigmatically2]

@Scillyjosh
I following the link to the Lidar that was posted by Mozzy, pulled up the spec data sheet and quoted that. So if you think the figures are wrong, tell the manufacturer, not me

 

[barfy]

I would think there may be advantages to capturing the sail shape data using trim stripes and optical capture? The capture regions can be better defined and real time solutions displayed. And I don't believe it is prohibited in racing.
I know it is already done, but surely advances can be made In this area?

 

[Scillyjosh]

@Scillyjosh
I following the link to the Lidar that was posted by Mozzy, pulled up the spec data sheet and quoted that. So if you think the figures are wrong, tell the manufacturer, not me

Sampling accuracy doesn't equate to resolved detail. I belive you quoted the angular sampling accuracy which is effectively just the lidars ability to know the angle at which a data point is taken - the actual resolved detail is as described in my post. This means you know where a datapoint is taken, but the gaps between data points are very large and as such you could be missing a lot of information. When it comes to signal processing you can do a lot more with more data!
The range resolution is 1mm but all this means is the data is reported to the nearest mm not that the data is accurate to the mm.
The data sheet is correct, the numbers you quoted just don't mean what you described.

 

[Scillyjosh]

I would think there may be advantages to capturing the sail shape data using trim stripes and optical capture? The capture regions can be better defined and real time solutions displayed. And I don't believe it is prohibited in racing.
I know it is already done, but surely advances can be made In this area?

Segmenting images is also very difficult and it becomes hard to get good data at the top of the sail. The advances in machine vision for robotics certainly have helped this field of sail shape measurement - libraries like opencv have certainly helped accelerate the entry point to basic implementations.

 

[shebeen]

Sampling accuracy doesn't equate to resolved detail. I belive you quoted the angular sampling accuracy which is effectively just the lidars ability to know the angle at which a data point is taken - the actual resolved detail is as described in my post. This means you know where a datapoint is taken, but the gaps between data points are very large and as such you could be missing a lot of information. When it comes to signal processing you can do a lot more with more data!
The range resolution is 1mm but all this means is the data is reported to the nearest mm not that the data is accurate to the mm.
The data sheet is correct, the numbers you quoted just don't mean what you described.

@Scillyjosh thank you for your insight, experience and incredibly straightforward explanations here. Much appreciated, and extremely interesting to me.

 

[dogwatch]

Banned for meteorological purposes. Not afaik banned as such.

 

[Stingray~]