Hammer Looking for a Nail, AI Drone Looking for Business Case

[maxwelllwang]

I’ve fulfilled my duty to Roko's basilisk. I made an Aerial Intelligence system, and I call it Dispatcher. Jarvis on a drone, skynet, or compute token eater, its basically a drone you don't have to fly because it understands and takes action based on the world around it.

Video of me “flying” the drone while I drive and rant about 3DR https://youtu.be/4MGjsrUMLZk

Watch it respond to an actual structure fire (DFR 3.0) https://youtu.be/v2dUJp2U_LY

Although its a regulatory nightmare I think this is what Drones as a First Responder will look like in 5-10 years. Officers have on-demand air support they can talk to like a police helicopter.

Quick shareable describing features https://youtu.be/DKVhVG7Fm84

I see many obstacles in the way to making this MVP into a product. There are plenty of use cases but no path to market. The liability and regulatory implications of a unmanned system making autonomous decisions while flying over people and properties seems insurmountable. It relies on off-the-shelf hardware that cannot compete with DJI’s. The drone has a lot of onboard compute and has unlimited range thanks to 5G, and the control interface is well abstracted so you could integrate any web UI, REST, mobile app, or even normal phone calling, but still relies on really beefy cloud compute for the Multi-Modal model. The current C2 UI is web based so I've been sending people links and they've been flying the drone from different timezones. I could see this as a very powerful integration into existing UAV systems such as Andruil’s Lattice, Skydio DFR command, Brinc’s DFR solution, etc; but as a standalone product it would have a difficult time finding PMF.

I’m looking for partnerships, potential integrations, or a co-founder to take this to market. If you have any interest in the project please reach out, I’m happy to give demos.

[threatofrain]

> Although its a regulatory nightmare I think this is what Drones as a First Responder will look like in 5-10 years. Officers have on-demand air support they can talk to like a police helicopter.

Hey, what you're talking about does have regulatory difficulties but they can be overcome with connections. You also need a different roadmap to success that is not just police and fire as there might not be enough money, and if you could you could compete with DJI's pricing and polish you'd have a different roadmap anyway. Also, early success in the US could help you push into opportunities with international customers.

Are you experienced in building relationships with police and fire? I'd start there as it's quite doable. They will help you cut through some FAA red tape but you'll also have to do some of the lifting for them too.

While the situation with DJI and Skydio may look very difficult to surmount, there are some directions you can take your drone that makes it more fit for fire and police use cases as well as beyond, you just have to focus on capability differentiation and not price and polish (DJI's Dock 2 pricing is amazing). For example, collision avoidance or high wind resistance. DJI is also facing uncertain political climate as Florida recently banned them for the use case you're talking about.

There are quite a few people working in your space, as you know, and drones as part of the first responder story are already becoming real in some cities.

Would very much love to chat and exchange ideas, shoot me an email at [email protected].

[polonbike]

If you're targeting first responders, you could approach private ones first (festival, big event, fairs, etc..). They have basically the same needs as the public ones, but their operation theatre is sometimes/often a private one (private property) with a lot less regulatory overhead.

Obligatory IANAL here, but (not US) event organiser.

[moonmagick]

This advice could probably apply to an interesting swath of new startups

[looofooo0]

Well if you can fly it to a certain area and crash it into any vehicle, drone or human detected with double digit probability, Ukraine probably would like to buy millions of them.

[echoangle]

I didn't look at all the videos but I'm assuming the actual computation is still done on the ground here. In this case, the utility is not as big as with on-device processing. If it were on the drone, it would be much more jamming-resistant (the only thing that could still be jammed would be GPS). But on-device computation would increase cost of course because every drone needs a bigger processor/GPU.

[bambax]

This. If it's completely autonomous and doesn't need either RC or GPS and can build its own route using geographic features on the ground, every nation at war will want it in huge quantities.

[throwaway48476]

This is the real market for this.

[flessner]

Great demo! The drone-space is really interesting at the moment. There's great hardware and open source software available to hobbyists.

Potential use cases are almost endless: agriculture, first responders, defense, maintenance, planning, transportation...

A lot of the value proposition will probably come from getting these drones to be as effective and autonomous as possible - this is easier to do if you have a domain-specific product.

As for myself, I am currently writing my own flight controller for Quadcopters in Rust targeting Linux and STM32s - quite the journey as I had no embedded programming experience before starting that project.

[GianFabien]

You could take Uber's approach. Ignore regulations. Focus on delivering such massive value that the regulations are changed to suit you.

[toss1]

That is a really good way to draw some crushing fines and law-enforcement attention; the FAA is not a local taxi committee, and they do take seriously the safety of bystanders. That said, if you want to take that route, build out ops in less-regulated countries. Iirc, that's part of the reason Zipline started scaling in Africa (along with greater need).

[Hizonner]

Sorry, did Uber actually deliver any value other than crashing wages, bargaining power, and working conditions for taxi drivers?

[edg5000]

In the 60s we could do a small amount of basic numerical compute with tubes, relays and crappy germanium transistors.

From day one, there were use cases for computers, as long as they provided tons of value with very little compute. It took until the late seventies for computers to be cheap and useful enough lower-value small-scale use cases.

Particularly banking could extract a lot of value from, say a 1 Hz CPU, because human computers more of a hassle to deal with than 9000 inefficient tubes or relays In the 60s, using computers to decode h256 video to watch some cooking instructions would be absurd.

Today, we extract value from drones, at a certain cost. The cost is determined by how much of a pain in the ass it is to use. For good PMF you need to get the value to pain ratio high enough.

These are the costs/pain points that hold back further adoption, in my personal experience:

Flight planning:

- Check adjacent land use (recreational, harbor, airport) for regulatory reasons.

- Check NOTAMS and airspace restrictions.

- Obtain airspace clearance (requires a phone call currently I believe)

- Check wheather, esp. wind speed.

- Ensure ground area is controlled and all people are informed.

During operation, continuous observation from operator required in order to detect the following potential issues:

- Battery level critical during RTH (time between low and critically low battery voltage can only be estimated and is affected by battery wear).

- Compass failure. State estimation algos will reject bad compass readings and fall back to GNSS for yaw estimation, so this can be avoided with good code)

- GNSS failure (can always happen due to tall obstacles)

The ground area needs to have no uninvolved persons (unless drone has safe abort system such as a parachute), because the following could cause an unplanned meeting with the ground:

- Battery failure

- Airframe failure

- Collision

- Thrust system failure (prop, motor, inverter)

- Flight controller failure (easy to lock up the MCU with some bad code)

If the above was not a reality, we would see more drones, esp with full autonomy. There are a lot of possible failures that are hard (but not impossible) to automatically detect and mitigate. Effectively constant monitoring by an operator is needed (this can be done remotely from a control room).

[threatofrain]

Here's some thoughts that applies weights to some of the items on your list.

- NOTAMS and flight restrictions should be automatically ingested.

- FAA restrictions are different for emergency services.

- You should have multiple drones in your fleet to deal with low battery problems.

- Collision detection is a hard problem and part of why Skydio got so much money.

- One point I don't see on your list is dock. IMO a dock is essential to play in the fully automated drone flight space, regardless if you're talking about site inspection or first responders.

[edg5000]

From operating drones in the field I noticed that it is needed to keep an eye on the vehicle, but also on the ground control station. I ideally keep my hands on the joystick in case I need to take over control from the computer.

Everything is automated, but supervision is needed and I need to be on the ready in the event of an emergency.

I was thinking about those VR raybans that would provide an overlay HUD. Combined with voice commands that could be quite something.

Essentially you are working on improving the UI side of drones, exploring different ways to interact with the drone. You have probably seen QGroundControl, Mission Planner and some of the proprietary apps. Essentially all these features could in theory be voice controlled.

I have personally never used voice assistance so I cannot comment what that would be like to use. The only thing I noticed in field work is that you need 8 hands and 8 eyes essentially, so voice UI could help there.

[o-o-]

Really cool engineering feat!

I think you're making your route to market overly complicated by calling it autonomous. Just call it what it is – a drone with a AI navigation system. As long as a person commands it and monitors it (and thus _is responsible_), there's no reason to call it autonomous.

First video has 131 views – I hope this blows up.

[MrHamburger]

AI drone with pattern recognition of military vehicles and personnel is being furiously developed on Ukrainian and Russian side to circumvent ever present electronic warfare. So the model is expected to run offline on cheap disposable hardware. That's where the most obvious business case today is.

[saaaaaam]

Straight away I can see use cases for this in construction inspection, surveying, facility management and so on.

e.g. “fly around this construction site to inspect these things from these angles and return to base” or “fly around the exterior of this building capturing detailed footage of these things and looking for any anomalies vs previous survey footage”.

I’ve been playing with high resolution satellite photography to identify anomalies on the rooftops of buildings we manage - eg plants growing near critical infrastructure on the roof such as solar panels or smoke vent equipment. Carrying out these inspections manually is often complex and expensive (two man crew, rope access, cherry pickers, etc) and although satellite photography is cheaper it’s still somewhat costly.

Another extension of this would be identifying abandoned vehicles on public streets - or in private car parks attached to offices or apartment buildings.

At a city/borough level use existing data to identify the areas that tend to be used to adandon vehicles and conduct autonomous fly-by on a regular schedule looking for vehicles and capturing the licence plate details. A vehicle hasn’t moved for 2/4/8 (whatever schedule matches known data) days, maybe not an issue, but check the licence plate and look up to ensure the vehicle is roadworthy and insured (I’m in the UK where we have specific databases for this).

If the vehicle hasn’t moved after eg 30 days, contact the owner checking everything is OK. If it’s not moved after then monitor or if other indicators (roadworthy/insured) change begin legal process. This is typically done manually by wardens here in London and because they often miss vehicles that are abandoned unless they are in controlled parking zones - but even then the vehicle might have a 12 month permit but have been abandoned. This is a much bigger problem on private land as often the parking enforcement people don’t visit regularly (or there may be no enforcement) and if the vehicle is abandoned the person doesn’t necessarily care - or may not be receiving - notifications about tickets.

Another use case could be public parks - carry out a regular survey, identify any anomalies present against previous survey and flag for manual inspection. You could have a small fleet of these drones which would carry out surveys before opening and after closing, each raised with specific surveys of areas, and on a schedule “check the net of the basketball hoop every third day” or “examine the gates of the tennis courts for signs of rust”.

[Yacovlewis]

Following this out of interest. I built an AI clone app for iOS, and faced a similar challenge around looking for a business case.

We are taking it into the visualization / interview prep space, where I think the tech offers a 10x improvement over a mirror or someone's imagination, instead of replacing another human, where it's 10^9X worse.

Andy Rachleff (Wealthfront, Benchmark) says venture-scale startups are about developing new tech and then finding business cases for them. It's helpful as founders to pattern match, watching how it's done in AI drones, AI iOS apps, etc.

[skanga]

Hi Maxwell, I did work on some similar stuff and have some thoughts, suggestions & connections that could help. Feel free to reach out to my yc username at googles email service.

[amrangaye]

Sidebar: interesting how you typed out the email address (I’ve also seen variations like me at gmail dot com). This worked for spam and phishing crawlers that just did a brute force search across texts. But now that we have LLMs, I’m wondering if this will be effective anymore, since it gets the context as well and is not just a “dumb” regexp search?

[Hizonner]

So in your video, the drone kept hovering over the Trader Joe's parking lot while you talked. What would it have done if you'd truly forgotten about it and not told it where to land?

[cpersona]

Real-time mesh networks for disaster recovery. Cameras would be a plus as well. Each drone can be a node in the redundant network. If they can pick up cell phone signals, then they can be even more useful.

I could see this being useful for ranchers/farmers. It can be used to identify loss.

It can be used to identify invasive species and where they are flourishing. Or to find where mosquitoes are breeding and possibly prevent them from doing so.

Some of these may not require regulatory approval.

[ForrestN]

Security companies? Seems like a lot of people would pay to have this on a dock in the back of their house checking their property every so often. A lot of camera systems are sold and this could replace a lot of permanently positioned cameras. Could respond to requests if something comes up, avoid being swatted down etc.

[vagrantJin]

Great shout, many security companies in South Africa have drone divisions, and drone companies that work primarily in security.

[Genego]

I really like this! I have been meaning to get into this space too. I have yet to touch any drone SDK, but I have found the possibilities of developing for and with drones in a lot of sectors quite interesting. Is there a way to contact you?

[Dicey84]

Maybe reach out to these guys (https://www.breakerindustries.com/)

Looks like you are walking down the same path in different places.

[simonebrunozzi]

Nice! Best way to reach out? Or, my email is in my HN profile.

[sarcasmatwork]

Very cool! I'm a drone pilot traveling Europe now. This could be very helpful for some of the shots I wanted to take.

How does it deal with other air traffic? NOTAMs etc?

[zooweemama]

This looks great! I really suck at making cinematic videos with my drone and I would welcome the possibility of just instructing it in natural language.