Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

DJI FPV

dji.com

Post reply on HN
Print view View original

111–120 of 162 posts

Re: DJI FPV

#111

Post by 05 »

StavrosK wrote: post #105
05 wrote: post #103

Earlier quoted context omitted.

It’s actually positive. Open source community has limited resources, so DJI proving this thing can be made to work well removes any uncertainty someone would have when starting their own project without that knowledge. It also affects the way people look at the feature - previously many would just automatically dismiss it as a useless gimmick, but DJI implementing it and people liking their implementation gives the i…

So the only reason nobody implemented digital video systems until now is that they weren't sure if it could be done, and DJI locking their stuff down is actually good for the hobby.

OpenHD. If you fly wings and 135ms latency doesn't bother you, you can build it yourself. Of course, it uses Raspberry Pi's proprietary Broadcom SoC h264 encoder, so much for open :)

Is TBS locking their proprietary protocol they use to differentiate themselves from commodity Semtech ICs good for the hobby?

Re: DJI FPV

#112

Post by bri3d »

latchkey wrote: post #84
wereHamster wrote: post #20

Earlier quoted context omitted.

High-res digital video has been available to the multicopter community for a long time now through https://www.dji.com/fpv . First released by DJI, later the technology was licensed to others (at least the camera / VTX side of the electronics), se for example caddex vista / nebula ( https://caddxfpv.com/collections/dji-1 ).

Interesting that it is all sold out too...

DJI seem to have diverted their supply chain from hobbyist FPV to this new drone. Supply for all DJI FPV gear has been very limited ever since they filed the "V2" goggles (for the Drone) with the FCC.

I'm hoping they keep their hobbyist/DIY FPV system going though, as it blows everything else out of the water. Analog video is well, bad, open source alternatives are either abandoned or mediocre, and the Divimath/Fat Shark system is not even comparable in terms of quality.

Re: DJI FPV

#113

Post by matsemann »

bee_rider wrote: post #102
matsemann wrote: post #13

The motion controller reminds me of a project I made many years ago, where head tilting of an oculus rift would control the cameras on your drone: https://news.ycombinator.com/item?id=7654141 Also cool that they have been embracing virtual practicing it seems. The original DJI controllers were impossible to pair with simulators, so I had to hack it: https://github.com/Matsemann/mDjiController/ I sold (or basically ga…

Does that produce VR sickness? Transmission time + going all the way to some motors seems like it would be pretty hard to fit in even 60fps, haha. I wonder if any consumer VR FPV drones would consider transmitting a 360 video signal (or just one that includes more FOV than can be displayed on the headset) to allow some head movements to be handled in software.

Yes, but not anymore than standard fpv goggles (with no tracking) would give in this case anyway.

Yes, we thought of that as a possible improvement ("future work"). It may be harder to get the stereo cameras to match up, is my guess. But for a single camera setup it's probably not too hard and already exists for 360 videoes. So I guess if one's capable of streaming the video it shouldn't be too hard.

And stereo cameras arent really that useful. We explored it, and found that when looking at stuff more than a few meters away it made no difference. For inspections (which was what our drone was made for) of close stuff it has some merit, but could still probably be handled by some IR sensor or other computational stuff adding an overlay to the video.

As for vr sickness, the most jarring experience was flying behind yourself like a 3rd person shooter haha.

Re: DJI FPV

#114

Post by bri3d »

dmitrygr wrote: post #15

50Mbit, 28ms latency wireless link at 12km distance, and on a battery budget, eh? I'd love to hear more, I love science fiction.

This has been out for a few years in the form of the "DJI FPV" system, Goggles and Air Unit, for DIY drone construction. It's not science fiction - the latency and bitrate taper off a bit at range, but it genuinely works.

If you tear down the hardware, it's using an Imagination IE1000 802.11ac frontend, Skyworks amps, and a custom DJI image processor chip (marked P1). So the air protocol probably looks somewhat like 802.11ac (OFDM etc.) but without the framing overhead and presumably without TCP/IP.

Re: DJI FPV

#115

Post by bri3d »

LukaD wrote: post #8
Dave_Rosenthal wrote: post #4

There has been a huge divide in the drone world between high performance first person view (FPV) drones capable of acrobatics and the "tripod in the sky" type drones that companies like DJI have popularized. The former has been very DIY and the latter have attracted companies to build integrated solutions. This new drone from DJI brings an truly staggering amount of innovation and integration to the FPV space: - Low…

> - Low res analog video -> high res digital video > - Modest transmission range -> long range Those are not really new. DJI has been selling their standalone digital FPV system for a while now. And there are lot's of different long range radio systems available. But I agree that this is impressive. I wonder if Betaflight, etc. will also implement some kind of bail out feature.

Betaflight already has very rudimentary RTH / "GPS Rescue" functionality, obviously without ground/forward sensors it's not as good, but it does work pretty well for "oh no something has gone wrong" in the form of video or control link loss.

Re: DJI FPV

#116

Post by 0_____0 »

joezydeco wrote: post #88
comboy wrote: post #26

Earlier quoted context omitted.

Ocusync has been out for some time now and it already sounds like magic. I think when compared to wifi or some radio bridges it usually operates in much less congested environment. Nonetheless, if you think about frequencies used, energy going down with r^2 and bandwidth they are doing, it's an amazing feat even in ideal conditions. And here's it's a moving object, on battery, with four electromagnets around it.

My son has a Mini2, the transmission range on this is pretty amazing. We were watching a neighbor's house over the holiday, which meant mostly checking for delivery boxes on their porch and putting them inside. We could drive the Mini2 out the window of our house and 2 blocks away to check the porch without getting out of our seats. And this is a suburban area full of houses, trees, and a shit ton of interference at…

Boasting about flying beyond line of sight without a spotter in a cluttered residential neighborhood isn't going to score you many points among the drone enthusiast crowd.

Re: DJI FPV

#117

Post by michaelt »

cromwellian wrote: post #52

Wait, why does the goggles look like VR goggles, but the drone only has 1 camera? Wouldn't this FPV system be way more awesome with binocular vision and head tracking? Imagine flying the drone, and moving your head would move the camera gimbals, and you'd experience it like you were in VR in the drone?

> Wouldn't this FPV system be way more awesome with binocular vision

Human binocular vision depth perception is actually extremely short range. It varies from person to person, but some measurements put it as short as 10 feet [1] - beyond that, the mind starts depending on other cues [2] like motion and relative size.

So the costs in bandwidth and hardware might not be worth it - getting your drone that close to things sounds like an expensive skill to learn :)

[1] http://visionlab.psych.sites.olt.ubc.ca/files/2016/03/131_Wa... [2] https://en.wikipedia.org/wiki/Depth_perception

Re: DJI FPV

#118

Post by eigenvalue »

05 wrote: post #110
eigenvalue wrote: post #99

The first-person view seems like it would be a total blast to try out-- basically every kid's dream toy. I would have given anything for something like that when I was 12 years old. But from a pragmatic standpoint of using the drone to capture video footage, I feel like having robust "self driving" capabilities (in particular, obstacle avoidance and smart target tracking) is more interesting. I was blown away recentl…

Note that Skydio isn't immune to thin branches - AFAIK no current tech can reliably detect them at distance sufficient for avoidance at high speeds. There are Youtube videos demonstrating Skydio crashes caused by these (and similar low visibility obstacles). Also note that while Skydio has technically superior avoidance tech (360 view vs forward only on DJI drones), it failed to capture non-negligible consumer market…

I’m surprised there aren’t lower tech ways of detecting thin branches besides doing real-time analysis of the live video feed. Like some kind of primitive proximity detection using radar to augment the image based approach, and if there is any doubt, taking quick evasive maneuvers.

Re: DJI FPV

#119

Post by joezydeco »

0_____0 wrote: post #116
joezydeco wrote: post #88

Earlier quoted context omitted.

My son has a Mini2, the transmission range on this is pretty amazing. We were watching a neighbor's house over the holiday, which meant mostly checking for delivery boxes on their porch and putting them inside. We could drive the Mini2 out the window of our house and 2 blocks away to check the porch without getting out of our seats. And this is a suburban area full of houses, trees, and a shit ton of interference at…

Boasting about flying beyond line of sight without a spotter in a cluttered residential neighborhood isn't going to score you many points among the drone enthusiast crowd.

Heh, true. Thanks.

Re: DJI FPV

#120

Post by kevinsundar »

jaywalk wrote: post #61
wepple wrote: post #50

I’d love for someone to build an overIP bridge for this, so I can sit on the couch at home and fly through the Grand Canyon. I’d happily pay a reasonable sum for it. I wonder if it could be done using a Stadia controller and chrome cast?

Who's going to let random Internet users fly their very expensive drone?

This was actually a thing a couple years ago. I tried it out a couple times and it worked really well: https://www.flythere.com/

Basically the gave you a big "box" of air space that had 0 obstacles that you could fly around in. It's impossible to crash the drone and it takes off / lands automatically.

Post reply on HN