[HN Gopher] Drone Racing League Creates End-to-End Cloud Production
___________________________________________________________________
Drone Racing League Creates End-to-End Cloud Production
Author : stikit
Score : 26 points
Date : 2021-05-17 14:26 UTC (8 hours ago)
(HTM) web link (www.sportsvideo.org)
(TXT) w3m dump (www.sportsvideo.org)
| cjlars wrote:
| I feel like the whole point of drone racing is to have that sort
| of video game in real life situation. You can have tighter turns,
| narrower gaps, and far more glorious crashes than you can get
| IRL... All without the risk to life and limb you see in, e.g. F1.
|
| Recreating the same experience virtually is just boring. Like
| they took a game idea, made it IRL -- and full of IRL design
| constraints -- and then took all those disadvantages back to the
| virtual world. I just have no idea what the league is thinking
| here.
| bri3d wrote:
| I think DRL know this and are looking to move to in-person as
| soon as they can, the article just focuses on their "cloud"
| move during the pandemic since that's what's what was
| interesting to the video production audience.
|
| The DRL Simulator is also... not great, even among drone
| simulators, so I suspect pilots who have been practicing in
| real life will appreciate the move back to real life as well.
| cjlars wrote:
| I think it's a cool concept and have been following it for a
| while, but it's worth noting that they started pushing the
| stimulator before, not after, the pandemic started. I think
| maybe to create a feeder league or something?
| bdamm wrote:
| Drone racing has a serious home-build streak in the actual racing
| robots. Kinda hard to virtualize that. If it's just a video game,
| what's the point? Maybe as its own racing class perhaps?
| Steltek wrote:
| DRL uses a standardized quad provided by DRL to the pilots. You
| have custom controllers and goggles but the craft are all the
| same. This is my recollection from a video by Nurk (DRL racer)
| anyway.
|
| And why wouldn't it work? A lot of Cycling turned to Zwift
| during the pandemic as races were cancelled. And E-sports have
| been a thing for a while now too.
| jcims wrote:
| Tangent warning lol.
|
| I just set up my first FPV drone a week ago and its an absolute
| blast. I've tinkered with RC trucks/planes/drones over the course
| of 20+ years but haven't put in enough time to get the natural
| perspective rotations required to drive them well.
|
| It's a bit of a nightmare trying to figure out what all you need
| (Josh Bardwell on YouTube and fpvknowitall.com was a _HUGE_ help)
| and getting everything talking well takes a day or two of
| tinkering, but it has been totally worth it. Strapping on those
| goggles makes an enormous difference. I still have a long way to
| go with stick coordination and just basic flying skills, but now
| it feels much more achievable and is way more fun.
| grahamburger wrote:
| Something I've been wondering about - are your goggles
| stereoscopic, and does the drone have two cameras? I've (very,
| very briefly) looked in to FPV stuff but quickly get lost in
| marketing stuff that doesn't seem to clearly indicate if
| stereoscopic equipment is available.
| bri3d wrote:
| No, almost all FPV pilots fly in 2D. Adding two cameras to an
| analog FPV setup would be trivial, but stereoscopic goggles
| are fairly non-existent, you'd have to try to produce a
| pipeline to get video into stereoscopic VR goggles without
| increasing latency to an unusable level - a very hard
| problem, really!
|
| With that said, I fly a lot of FPV (planes and wings) and
| I've never felt wanting for true stereoscopic 3D. I have
| flown a simulator on Oculus goggles in stereo and the
| experience wasn't that different - I think there's enough
| movement / visual cue that your brain is good at
| interpolating the 3D aspects anyway.
| difosfor wrote:
| I would imagine you could quite efficiently encode a second
| view that's largely the same as the first You'd need more
| processing power, but not much more bandwidth that way. But
| the VR 180 videos I've seen so far just seem to do side by
| side encoding. I guess you need hardware accelerated encoding
| and decoding for practical use so that's preventing such a
| new codec for now. Unless the side by side approach is
| already efficient at this
| jcims wrote:
| There are stereo lens adapters that might get the job done
| without any additional electronics. Resolution would suffer
| but the extra spatial acuity might make up for it.
| jcims wrote:
| No stereo with the DJI goggles but there are some that do
| support it. There is enough reference info in the relative
| motion of objects while you are moving to create a 3D space
| in your mind. This instantly evaporates when you stop moving,
| however, particularly when near foliage.
| bittercynic wrote:
| I haven't seen anything off-the-shelf, but I think this would
| be a really fun thing to set up. Maybe with a wider camera
| spacing to get some depth perception even at longer
| distances.
| TooSmugToFail wrote:
| Vast majority of the FPV systems out there are indeed
| monoscopic. Stereoscopy is too resource-intensive for FPV,
| and it never caught on.
|
| The only viable digital system out there, DJI's Air Unit and
| their Goggles, is really pushing the boundaries of what's
| possible with available silicon, and they can only do a
| single-channel 720p low latency video stream over the air.
|
| Attempting to do digital FPV in stereoscopy would require
| additional compute on the encoding stage, and an additional
| Tx on the air unit. The AirUnit is rather chunky as it is, so
| having something that can do 2 simultaneous 720p channels
| would hardly be viable for a 5" drone, let alone 3".
|
| It would also use extra radio channels too, halving the
| number of available channels (determining how many pilots can
| fly simultaneously; currently at a max of 8 - at least
| theoretically).
|
| The rest of video systems on the FPV market are still analog
| (in PAL/NTSC).
|
| Interestingly enough, a few years ago a rather obscure
| Chinese FPV brand Skyzone had goggles which supported a
| stereo FPV camera block.
|
| Basically, their rig had two cameras, each transmitting over
| its dedicated video Tx, and the goggles were equipped with
| two video Rx (VRx) units, each capable displaying video on a
| separate goggle display (left/right eyepiece). Most analog
| VRx units used for FPV today have two separate receivers, but
| they are used in 'diversity' mode, where they are tuned to
| the same channel. Skyzone also had twin Rx setup, but they
| could do both diversity and stereo.
|
| Thinking of it today, it was an amazing piece of speculative
| innovation that FPV is rife with, but as I said: it never
| caught on.
|
| People that used it were like 'wow, it's different...' but I
| never saw anyone do a build with that for something more than
| a 'Hey look: stereo FPV' type of YouTube videos.
|
| My guess is: you need to get rather close to stuff in order
| to be able to perceive depth afforded by a stereoscopic
| system with cameras set at a distance small enough to fit on
| an FPV drone. Possibly a bit too close, for an FPV drone. So
| it never really made sense I guess, and so it never caught
| on...
| snypher wrote:
| Usually just one camera. I'm not sure if you could mount two
| far enough apart to get a good angle for a 3D tree.
| devteambravo wrote:
| Yeah, this seems like a great idea. Let's combine gambling with
| video games kids play..
| moshmosh wrote:
| Gambling is already a huge part of esports. Huge. Getting a
| glimpse into that world helped me make sense of the whole
| phenomenon. (as in: oh _that 's_ why there's so much money in
| esports)
| vmception wrote:
| That's not a strong argument to support this or any stance.
|
| Watching the thread to see if different arguments or
| observations appear.
| skizm wrote:
| Sites like loot.bet have been around for a while now.
| canadianfella wrote:
| What's the issue?
___________________________________________________________________
(page generated 2021-05-17 23:02 UTC)