Hexo+: Autonomous Aerial Camera | Hacker News Reader

Hexo+: Autonomous Aerial Camera

kickstarter.com

68 pointsdanboarder30 comments

I've spent the past year working on a similar concept using the same flight controller, but didn't think that its safe enough for everyday consumer use, not at least for another couple of years until radar or optical obstacle avoidance is available in the consumer space. Everyone who operates MAVs knows that things are guaranteed to occasionally go wrong even in ideal conditions. If GPS or INS glitches, the drone becomes a high velocity flying circular saw - and make no mistake, those flimsy plastic propellers will casually cut you to the bone at decent speeds.

I've had a number of aborted flights and close calls during development, the scariest one being a GPS glitch while testing full autonomy on a crowded beach. The most common issue is poor GPS accuracy which can cause autonomous flight to be erratic or completely uncontrolled. In these cases, the operator will notice that the drone isn't responding correctly and switch to manual control to recover. The worst possible issue is a GPS failure or occlusion: if the system loses GPS during autonomous flight, it has about 10 seconds of guessing its velocity based on other sensors before it becomes completely out of control. In these circumstances, again the operator will quickly take over to recover. GPS dropouts or poor accuracy are a regular occurrence in suburbs and guaranteed in built up cities at altitudes below the skyscrapers. The only time it's more or less reliable is over water and rural land.

I think giving drones without a means for manual control to untrained consumers is almost negligent and a disservice to the industry. Speaking about the 3DR flight controller and software in particular, there are a number of known issues such as the dreaded uncontrolled flyaway which are guaranteed to happen to one of the users at some point. God forbid one of these flies into a kid and leaves them disabled; the FAA is just waiting for an excuse to crack down on hobby MAV use.

...the drone becomes a high velocity flying circular saw...

How would efficiency be affected if the propellers were surrounded with wire shields, like a household fan might be?

I don't know about wire shields, but circular guards acting as ducts should theoretically increase efficiency for quads due to the relatively slow propeller speeds. The downside is they add mass and I've yet to see a compact folding design that incorporates guards.

I think we'll see guards become a very common thing as designers find clever ways to fold them in the near future. I also think that that FAA regulations for low altitude drones in the future may very well mandate guards because they go a long way towards preventing injury.

Ducts around the fans might be prohibitive from a drag in forward-flight point of view, even if they are desirable from a static hovering efficiency point of view.

I don't actually have the numbers on that, but it's probably something to look out for.

It depends on how you're flying but drag is generally ignored for hobby use at this stage. Increased drag will affect top speed and range, but endurance won't suffer much because the vast majority of energy is used for altitude control. If you reduced drag by 99%, the improvement in endurance would be modest for the same reason.

Its a much bigger consideration for point to point missions like cargo transport though.

What is the plan for your project?

The team section lists 3 "development" staff: there's a CTO and two developers. There are 6 people on the team doing design, videography, and marketing, including the CEO who does list himself as a "telecom and electronics engineer", and so might be involved in design.

A startup building hardware - not to mention advanced image-tracking, autonomous navigation, (hopefully) hazard avoidance, which are hard problems - would in my estimation have more staff doing engineering and manufacturing, rather than design and sales.

Is this a reasonable team arrangement? Am I underestimating the skills of their technical staff, are they outsourcing engineering, or are they using a preexisting plaform? Or am I underestimating the importance of design and sales (which have certainly produced and brought to my attention a gorgeous Kickstarter page)?

From the Kickstarter description, they're leveraging existing open source software: "The onboard software is based on the 3D Robotics open-source code"

Also note, they DO NOT include hazard avoidance in the current product: "The reason for this is that there is no avoidance system included in the first version."

So I think your questions are legitimate. The underlying question is: do we now live in an era when a team of mostly non-technical people with a couple of smart hackers can put this kind of product together from existing technologies without inventing anything really new? I guess the answer is mostly yes but I wonder if obstacle avoidance is really something that can be left out without serious consequences.

I dunno .. does this really need yet another drone body and rig just for the sake of the product? I mean, wouldn't it have been a little more sensible to add "follow-me" features to existing drones - like, this could just be another feature for ArduPilot .. does it really require all the product hype to replace this? I think this is sort of a 'me too' factor with these kickstarter campaigns - whereas in the past a hacker would just add the feature to the existing open code, now its become a reason to start a whole new company and get funding up front, first.

I think, rather than spend any money on these guys, it'd be simpler/less problematic to just add follow-me to Ardupilot. With a little bit of bluetooth/wifi tracking magic, maybe it'd be easy, who knows ..

Your suggestion to add this to an existing unit might well be the idea behind their cheap "3D print" option. They won't suggest that we do this, because they don't want to support the integration, but it looks like they aren't really trying to prevent it.

How does it compare to AirDog: World's First Auto-follow Action Sports Drone (https://news.ycombinator.com/item?id=7897776)

It doesn't. Because it looks like HEXO's drone in the videos is a fake =]

Basically, they have a mockup and a campaign which started few hours before Airdog's.

It's much less expensive, for one thing.

Easy to be less expensive when Hexo does not have a working prototype and are still in the hype stage.

Kind of sad though that Airdog is going to get less funding(and less exposure) than Hexo, just because of lack of flashy marketing.

Still, looks like both projects will get funded and that is a good thing.

A while ago my wife and I were having a debate about hollywood. Mostly about how much it sucks, and why. I made the statement, it doesn't really matter because technology is bringing down the cost of making a movie for everyone in literally every way. At the time I was thinking about cheap CG, and DSLR's. Great quality filming like aerial shots were one of the few holdouts that individuals didn't have access to.

Now one of the last few holdouts are good quality talent, but I think people can learn to improve skills, and technology can help there too.

To me there isn't a single "hollywood killer" but rather its a singularity of sorts. When technology converges to the point where the quality of million dollar marketing oriented mass audience films are irrelevant. Because cheaper "niche" movies are more compelling anyways.

Great narrative storytelling talent isn't going to expand much. As technology reduces the costs of everything else, the Spielbergs and DiCaprios will capture all of the profit in the market.

Cool, but I thought "follow-me" technology was already pretty advanced. Does this do anything different?

I'm a rower and a few people are looking at using follow-me drones for rowing training and coaching. Here's one example (not follow-me, but shows observation from a drone): https://www.youtube.com/watch?v=0XfL_2EzEHI

The theory is that a drone follows the rower, maybe moving around the subject for different viewpoints, and the (possibly remote) coach can examine the footage afterwards. A coach can advise many rowers without needing to be on hand (or even in the same continent) for all of them.

Does anyone understand how it tracks the subject, i.e. the person being filmed? They mention the MAVLink protocol (http://qgroundcontrol.org/mavlink/start) but that doesn't seem like an answer.

Not sure if they have enough on-board computing power to do real-time image analysis and track that way.

Just guessing: it follows your phone?

How, technologically speaking? GPS is way too coarse for what they're showing.

Dead reckoning of phone trajectory using inertial sensor (of the phone) + gps corrections ?

I don't see a dedicated camera on the drone (using gopro for machine vision is not ideal), nor a beefy cpu to handle person tracking onboard, but I could be mistaken.

You can get centimeter accuracy with gps. http://swift-nav.com/piksi.html Even with cheaper modules and phones that dont carrier phase rtk, you can get meter which would be plenty for this application.

Phone sends GPS coordinates to drone.

I'm pretty sure they are doing image tracking. They speak about being able to frame the image you want before, and the drone keeping that framed:

We have designed our drone specifically for action sport, on the tracking mode we offer framing is more than just follow me: it allows you to set drone position relative to you (altitude, distance, orientation), so that you can get perfect aerial shots.

This seems like a really cool product if you're above the treeline or on the ocean. Otherwise, I'd worry about not having enough tree and lightpost -disposable go-pros.

If you suddenly pull the brakes of the motorcycles, will the blades cut your head off? I would worry about that before putting this out..

Having inadvertently put my finger in a Parrot AR2 drone's blades, I strongly doubt decapitation would occur. "Ouch, that stung a bit" is more likely.

http://www.rcgroups.com/forums/showthread.php?t=1144377&page...

RC helicopters have killed. A Parrot is a pretty small quad. I would find it surprising if they have liability insurance for this.

This looks much more comparable to a Parrot than to the deadly helicopter you reference. In general, do manufacturers carry insurance for claims stemming from incidents like sticking one's fingers into whirling blades?

No, the parrot is very lightweight http://ardrone2.parrot.com/ardrone-2/specifications/

And yes, it isn't the act of sticking the fingers into blades, it is its stated use of autonomously filming around a person. When I direct the camera to maintain an angle and elevation with respect to a target and it impacts that target, this is a liability problem.

Reply on news.ycombinator.com