Remind me, what’s the typical error in consumer GPS in the real world?
It uses a absurd time warping/dilation system, basically, given a known mountain bike course, and your activity on said course, I can use a solver I developed to “warp” your activity the course perfectly, it’s a different concept entirely than Strava’s segments that then lets me break down very minute portions.
I guess a better way to put it is, I can account for the error by figuring out when a cyclist exactly crossed the course in a know area multiple times, actually, thousands of times, to correct at a high degree their “error”.
I didn’t know this would have this must interest… perhaps I should finish that project…
Speed it up so it could be used in real time in esports so that bike and position on it affects performance/results.
I didn’t realize this would garner decent attention… because Strava is a jerk enough to patent the very concept of “creating a leaderboard for start/end of a segment from an activity gpx file” I had to get very creative, and ended up making something pretty ridiculous.
We even bought the domain https://truesegments.com for it (it’s still WIP with demo pages on the site and the tool isn’t open to the public, it needs refinement).
in fact, this whole wind tunnel mimicking idea was born because of that tangent. First, I created probably the most advanced segment system imaginable for mountain biking, because I was fed up with Strava’s segments being pretty inaccurate, here’s a screenshot, yes we’re going to make it look better:
Then I realized I could forgo the concept of predefined “segments” entirely, and just figure out where I’ve been improving on the same course, using this data to help figure out what tire pressure, bike combo, etc. is fastest.
I then also found out how to cut the fastest 5m for every portion of a mountain bike course I’ve been on from all of my attempts, cutting it together into my “theoretical fastest” time… and… lol… even though Strava claims to be using AI and whatever to screen KOMs, managed to take all of them for a course with regular XC races using the cut together activity:
Don’t worry, I deleted it immediately.
It soon became clear that I also wanted to know where I was faster when with what setup, so, I my twin went ahead and go-proed the route, I spliced up the frames and made a “Google Map’s street view” of the course.
Then, I wanted to be able to “click through” into the image, to advance forward, similar to Google, so, I started working systems to do this, accidentally manually writing scripts that were part of a field called “Photogramtry”, which I probably mispelled.
Then, I started exploring said field and started rebuilding the mountain bike course, this was one of my first attempts, from a Gopro video:
then I started getting better, this isn’t real:
Then I started building… the entire course.
Feel free to check it out:
That’s a portion, I wouldn’t recommend going their on mobile, and yeah, it loads slowly and has funky controls, and this is where I started running into problems, I couldn’t host the darn thing!
So, in efforts to reduce the size I made custom webGL viewers, custom shaders, custom polygon generation algos:
and eventually started scanning all sorts of things, here’s my Twin:
and started dabbling in Aeroscience, and, here we are.
1 Like
Very nice. I’ve “rebuilt” road courses using speed and power, and compared their virtual elevation over repeated runs over the same locations, but I don’t usually work with off-road segments. However, I was told by the Swiss team after the 2016 Olympics that they’d used VE to model the DH runs for Nino Schurter and to test various tunings. That worked out pretty well for him. I’ve also used a variant of VE to spot a suspiciously oddball data file from someone who’d made some unusual performance gains. That was very sad and depressing. I didn’t get interested in this area for that.
1 Like
depends on your interpretation of “guess”, it is a solver that uses thousands of intersection points to solve for a truer “where you were when”, but it’s neither here nor there, I care about that Aero magic currently.
Ha ha, I didn’t even know this was a ‘market’! When I started coding LowCdA as a personal project, I was just trying to solve my own training issues after a brutal 200k where my traps were so shot that I had to stay on the tops for the last 2 hours… might as well have been dragging a parachute!
Your approach with high-res 3D modeling and CFD analysis is exactly what’s needed for optimizing equipment. Trying to estimate actual drag coefficients from just a front camera view seems questionable to me - there’s so much more that affects drag than just frontal area
What I’m doing is much simpler - just giving real-time feedback during indoor training to help build better position habits. Although you’re spot-on about the orthographic projection challenge - I actually started with my iPhone on a tripod across the room before figuring out how to make it work at close range with a webcam. But at the end of the day, it just needs to help prevent getting sloppy during long trainer sessions so those good habits carry over to outdoor rides.
That said I think you’re onto something huge with the equipment testing angle, especially for triathletes and time trialists. Being able to accurately compare before/after changes to equipment setup through CFD would be invaluable - even with longer processing times, it’s way better than field testing where conditions are never consistent.
1 Like
yeah, and I can even rig the model and try a bunch of position changes very quickly, without having to re-create it or anything…
1 Like
Kudos…you need the right idea at the right time and we have a next David Tinker here (intervals.icu).
Your cervelo model above has 2 front cables. Could you remove those cables in the model and show us the difference in cda or watts saved at 30 km/h for example. Could you also do that for 4 cables vs 2 cables vs 0 cables?
That would be very interesting.
2 Likes
Hmm… that’s a good idea, and would take very little time/effort… im working on a technique for a higher quality model right now, when it’s done, and I have that, I’ll probably do that.
2 Likes
Do you have any news about the work on this?
1 Like