Slalom Ski (Handle) Path Physics part 4

[lundberg]

This is the 4th update for my effort to quantitatively understand what is going on while we ski. 

In this installment I've further refined the data processing and presentation, animated and time aligned it with video data.  It turns out that you can easily parse GPS, accelerometer and gyroscope data from the .mp4 files generated from a GoPro.  So I replaced my phone with a GoPro and went to the lake and skied. 

Assumptions:

1) Boat speed is correct and constant and path straight and centered (I could try and combine the GPS data "at the pylon" but I don't think it would add much value)

2) Rope length is known and there is no slack (the data is not as accurate in the turn when the rope has little tension in it)

3) The actual buoys are accurately placed (model assumes perfect locations)

4) The data shows the "handle" location and speed not the ski.

Here are 3 passes I skied at the same speed with different rope lengths (I don't claim to be a great skier 🙂 ).

The plot on the top right shows the "virtual" drone view of all three passes, and the lower two plots show the skier speed over the water and angle of travel (0 degrees = down the lake and +-90 = across the lake).  The bold blue lines and the beginning and end are the gates and the thin lines are buoys 1-6.

 

 

 

Here is a half speed version (single stepping the video is interesting too)

 

 

Here are three passes from a friend all at the same speed and line length

 

 

I'm curious:

1) Is this type of data and view is interesting?

2) How would you use it?

3) What other metrics would be valuable?

4) Are there other ways to show the information?

 

 

"Slalom Ski (Handle) Path Physics" History below,

It started in the winter with a parameterized matlab model of the handle path based on a swing profile.

part 1-

 

 

followed by a simple setup to measure the actual swing profile with rope tracking camera mount and phone app to record the gyroscope data

part 2-

 

followed by refined processing and actual data sets from the different skiers in the course

part 3-

 

 

 

[Golfguy]

1

[Adam Caldwell]

I think you need to plot down-course speed and cross course speed independently, which you should be able to do with your setup.  

Also, plot angular acceleration and speed.  that's what really changes as the rope lengths change.

I've thought it would be cool to get a Power or Watt output during events.  Would show a dramatic delta between passes.  They cycling world has a lot of cool gear/tech to capture that off the crank.  I think it would be a cool metric to add to the "strongest pull" they're doing at pro events. 

[Andre]

My head is spinning...

[503Kento]

Super cool, following with interest. 

[DvarianDan Johnson]

If this cannot be crystallized down into practical application - I see it as no value. It’s just stats. If that’s the goal - then fabulous, but again if this is to be used somehow it must be clarified 

[lundberg]

@Adam Caldwell you are right all the data is there.  In making and checking the following plots a found another minor correction to my processing (the zero in the integrated gyroscope data was not in the right spot).  Take a look at these plots and let me know if it matches your intuition (do the amplitudes between rope lengths make sense, does the timing / location of the peaks make sense....).  I can animate and align this data to the videos as well.

@Horton I'd like to replace the video clips in the first post, is there a way for me to do that?

 

Corrected video 

 

[lundberg]

@DvarianDan Johnson I can think of a handful of uses.

1) Coaching: Having actual data to compare or track is very valuable.  This would let you analyze your gates, how high / wide you are getting, how consistent your timing is (starting your move out .... when and how you turn in).   You can understand how early or late you are during a pass and what might have lead to that.  You could measure / identify what changes in actual dynamics when you adjust technique.  Compare your skiing to someone else's.  Measure progress even when buoy count doesn't change.  

2) Correct and standardize terminology and dispel myths:  "this ski is really fast / slow", "this ski get's more angle", " the pros go 90mph", "there is only one 6in wide ski path to run 41off"

3) Enhance webcasts: Embedded graphics and overlays in video replays between passes.  Compare skier vs skier on the same passes.  I've been watching Olympic swimming and they are overlaying swimmer speed to let the view notice things on their own.  Or like @Adam Caldwell suggested, find other impressive metrics to add to the competition.  

4) Research, Development and Marketing: Use the data to add objective measures to the subjective opinions of testers. Find new and innovative features or technology to improve equipment performance.   

 

I'm sure there are other uses; likely enhanced by smart people (or AI) finding the specific measure to answer a specific question.  Data is the key.

[DvarianDan Johnson]

@lundberg Data is great. I live in it daily. This however is a long way yet from  THESE RESULTS mean you should DO THIS, which is what end users want. More coffee for the lab…………..

[ToddL]

Before the data can provide meaning, there must be sufficient instances of capture to average out irregularities and each instances must be categorized into "good" (optimal) or bad" (sub-optimal).  After that, we would have an objective measurement of a skier's performance compared to a declared optimal performance at a given line length.   The patterns that will emerge within the sub-optimal instances will allow for scenarios of error.  Each scenario of error could then be analyzed for root cause(s).  Then, there would be an opportunity for corrective changes (coaching/improvements).   Further, with sufficient data capture of optimal passes at each line length and speed, other patterns and trends may immerge to help uncover exactly what is different about how they occur.  

Lastly, specific coaching solutions could be captured to see which elements have the greatest impact upon moving toward the optimal measurement.   

Or, we could just grab a few beers and talk about handle control, stack, etc. 

[ETskier]

I like your last suggestion best!  My old brain can't handle the data explosion.

[lundberg]

@ToddL one of my goals was to make the data capture simple and cheap.  That way getting more data is more realistic.  Unfortunately, I've only had one other person participate, @pkulio. 

The phone app I started with met the cheap and universal bill but the having time stamped video and sensor data from a gopro is much better.

Maybe, this example will inspire more people to dig out their GoPros.

Edited August 3 by lundberg

[503Kento]

@lundberg what do you need to contribute, I use a GoPro (7 or 10) to record almost every.set. 

[pkulio]

@lundberg This is really cool update! I don't have a good way to mount up my gopro so I'll play around with different sensor logging apps to see if any export with timestamps.  Hoping I can find a free version but willing to sacrifice a few dollars for the cause.

I think it would be incredible to have this info shown at tournaments.  If nothing else, its a great way to quantify how early or late someone is in the pass.  After one my passes, Andrew was able to create a spreadsheet of the down course percentage after each buoy.  Super cool!

Another awesome application that he hasn't brought up yet is for jump.  Here's some plots from a set I took back in Jan.  The first one is a 3/4 cut and pass, next two are double cut and passes, and the last two are singles.  Right off the bat, its valuable to see estimated speed at the turn, max swing speed into the ramp, and how early/late the cut is.

[lundberg]

@503Kento basic idea is to track the angle of the rope over time, so if your gopro videos are from a pylon mount that is moved by the rope it should work (some camera mounts have some slop (hysteresis) built in to make the video more smoother, this actually makes the sensor data less accurate for this application).  If you post a picture (or send a message to me) we can see if it will work.  Then if you share the files with me I can process and align the data with your passes.

If people are curious to see the raw sensor data from their GoPro videos you can use the free version of this tool (https://goprotelemetryextractor.com/free/) to view it and download .csv files of each sensor.

 

@pkulio what I found to be most helpful in these latest datasets was finding the precise time from the video when the skier is directly behind the boat at the gate and a rough offset on the gate timing (early, centered or late) and then use that to calibrate the sensor data.  Small errors can add up and this made for much less fiddling with the data.

[503Kento]

@lundberg I use a ski-doc, I’ll capture some video tonight and connect via PM on getting you the files. 

[503Kento]

@lundberg

I can't seem to send you a PM??

[lundberg]

Big thank you to @503Kento and @pkulio for sending me more ski data!  

Here is set from Paul 36mph 28, 32 and 35 off (slowmo) and his comments after looking at the data.

• Looks like I pulled out equally as high on the boat at 32 as I did 28, naturally causing a narrower gate at 32.  35 looks higher on the boat and the same width as 32.  This isn't super obvious from just looking at the video.
• Also pulled out latest at 32.  Maybe good thing I was narrower.
• Horrific pre-turn /  release into 1 ball at  the first 35 actually shows up crystal clear on the handle path.
• Big spike in speed out of 2 ball at 35 got me back into that pass.  Didn't shut down speed as quickly as the other line lengths, but did end up having a lower min.
• Interesting handle movements at 28 near the buoy - specifically 3 and 4 ball.  My guess is that that pass actually had a pretty tight line and so those movements out are due to the line coming tight through the turn.
• General trend shows lower maxes and higher mins in speed going through the course

 

then a bonus of 35 and 38 off

 

Here is some data from @503Kento @ 32.3mph 22off and 28off