Hi all,

I’m working on a project to build a drone that can detect potholes and other road surface issues. My plan is for it to fly at a distance of 15-20 km from the operator and detect these problems using onboard processing. I know I’ll need a good camera, a capable processor, and a reliable communication system, but I’m not entirely sure what specifications would be best. I’m also worried I might be missing some important details that could impact performance down the line.

I want to make sure I don’t overspend but also don’t want to use components that will limit performance as the project evolves. Any guidance to point me in the right direction would be really helpful! Here are the specifications I’m thinking of (though I realize some might be a bit ambitious):

1. Range: Ideally, the drone should work at up to 30 km from the operator and fly at altitudes of 4–6 km.
2. Camera: A camera that can spot road defects from about 1 km altitude would be great, but I’m willing to lower the drone closer to the ground if necessary. It would also be a bonus if it had infrared (IR) capabilities for low-light or night-time use.
3. Battery Life: A flight time of around 5-6 hours to cover a large area in one go.
4. Processing Power: The drone should be able to handle real-time image processing onboard (maybe with a GPU?) and create a 2D map marking detected road issues.
5. Environmental Tolerance: It needs to handle temperatures between -10°C to 50°C (14°F to 122°F).

I’d appreciate any suggestions on components I should consider or if there’s anything I might be overlooking. Thanks in advance!

 

Not at all practical for an absolutely absurd multitude of different reasons.
My advice is to forget about this idea entirely.
.
.
.

 

Not at all practical....

You're far too negative. The NSA can detect anomalies at that scale from space. It's just a matter of budget and manpower, which has not been specified.

I have little background in such matters, but surely a small team could build such a device for less than a few $million? Much less, if adaptation of existing commercial drones was acceptable. Then it's mainly an exercise in AI vision.

 

Your profile says that you are a student. This sounds like an initial proposal idea for a senior design project or something similar.

Where is this project idea and goals coming from? That will be a major factor in deciding what you do from here.

 

Frim your soecfication;

1. Range: Ideally, the drone should work at up to 30 km from the operator and fly at altitudes of 4–6 km.

From FAA regulations:

The maximum allowable altitude is 400 feet above the ground, and higher
if your drone remains within 400 feet of a structure.
…
you must still keep your unmanned aircraft close enough to be able to see it if
something unexpected happens.

 

This idea amounts to multiple Law-Suits just looking for a place to happen,
and possibly Prison-Time as well.

It would be much more practical to launch your own private Satellite-System.

A manually-operated Cell-Phone-App could provide GPS-Coordinates to an automated-dispatch-system.
There, problem solved.
.
.
.

 

Altitude of 4 to 6 Km, good to make detection of your drone, not so easy, eh?

 

Hi all,

I’m working on a project to build a drone that can detect potholes and other road surface issues. My plan is for it to fly at a distance of 15-20 km from the operator and detect these problems using onboard processing. I know I’ll need a good camera, a capable processor, and a reliable communication system, but I’m not entirely sure what specifications would be best. I’m also worried I might be missing some important details that could impact performance down the line.

I want to make sure I don’t overspend but also don’t want to use components that will limit performance as the project evolves. Any guidance to point me in the right direction would be really helpful! Here are the specifications I’m thinking of (though I realize some might be a bit ambitious):

1. Range: Ideally, the drone should work at up to 30 km from the operator and fly at altitudes of 4–6 km.
2. Camera: A camera that can spot road defects from about 1 km altitude would be great, but I’m willing to lower the drone closer to the ground if necessary. It would also be a bonus if it had infrared (IR) capabilities for low-light or night-time use.
3. Battery Life: A flight time of around 5-6 hours to cover a large area in one go.
4. Processing Power: The drone should be able to handle real-time image processing onboard (maybe with a GPU?) and create a 2D map marking detected road issues.
5. Environmental Tolerance: It needs to handle temperatures between -10°C to 50°C (14°F to 122°F).

I’d appreciate any suggestions on components I should consider or if there’s anything I might be overlooking. Thanks in advance!

Your specifications for altitude and flight time put this into the category of a surveillance drone. It is practically impossible for you to do this. If it just a thought experiment then research military surveillance drones to get an idea of cost and operational requirements.

The flight time issue is made worse by the high demand for compute that you include which will consume a very large amount of power. It is not clear why you want to do the processing on-board.

 

@LowQCab @WBahn @BobTPH I am not from the US. This project is a competition project by my city. So there are no lawsuit issues here.

 

""A manually-operated Cell-Phone-App could provide GPS-Coordinates to an automated-dispatch-system.
There, problem solved.""

And it's dirt-cheap, with virtually zero maintenance, and it's also virtually "false-report" free,
and can use existing technologies, very simply adapted to the particular requirements of the job.

Maintaining and Operating a large, sophisticated Drone is a very expensive endeavor.

The detection scheme is extremely expensive because it's all custom designed,
and therefore will require proffesional development and maintenance which would be brutally expensive.

Maybe in your Country there are no consequences for personal bodily damage, or property damages.

All of this is a MASSIVE pain in the back-side, not to mention the expense.

And what do You get ????
Bragging-rights ???,
look there, up in the sky, "isn't that cool" ?
but certainly NOT increased efficiency in maintaining the roads.
.
.
.

 

Your specifications for altitude and flight time put this into the category of a surveillance drone. It is practically impossible for you to do this. If it just a thought experiment then research military surveillance drones to get an idea of cost and operational requirements.

The flight time issue is made worse by the high demand for compute that you include which will consume a very large amount of power. It is not clear why you want to do the processing on-board.

Regarding on-board computation I just thought it would make it a more "compact" solution which can relay the info about the detected pot-holes directly to the operator without having to develop another system in between.
However, I do understand it is better to not do the computing on-board so as to ease the constraints on the drone, especially if the on-board computation makes the solution impractical. The logical alternatives seem to be live-streaming to the operator and using that feed to do the computations or recording videos on-board and then using them later on. The former would still use a lot of power(?) because of the distance from the operator(?) while for the latter the drone has to fly all the way back to the operator anytime a computation needs to be done making it a little less dynamic imo.

 

""A manually-operated Cell-Phone-App could provide GPS-Coordinates to an automated-dispatch-system.
There, problem solved.""

And it's dirt-cheap, with virtually zero maintenance, and it's also virtually "false-report" free,
and can use existing technologies, very simply adapted to the particular requirements of the job.

Maintaining and Operating a large, sophisticated Drone is a very expensive endeavor.

The detection scheme is extremely expensive because it's all custom designed,
and therefore will require proffesional development and maintenance which would be brutally expensive.

Maybe in your Country there are no consequences for personal bodily damage, or property damages.

All of this is a MASSIVE pain in the back-side, not to mention the expense.

And what do You get ????
Bragging-rights ???,
look there, up in the sky, "isn't that cool" ?
but certainly NOT increased efficiency in maintaining the roads.
.
.
.

Yes. I do get your point. It makes for a much simpler solution. But actually, they are looking for something this complex. They do need a drone. And although it does seem like using a fire hose to water a potted plant, maybe they want to build on this system later on.

 

Regarding on-board computation I just thought it would make it a more "compact" solution which can relay the info about the detected pot-holes directly to the operator without having to develop another system in between.
However, I do understand it is better to not do the computing on-board so as to ease the constraints on the drone, especially if the on-board computation makes the solution impractical. The logical alternatives seem to be live-streaming to the operator and using that feed to do the computations or recording videos on-board and then using them later on. The former would still use a lot of power(?) because of the distance from the operator(?) while for the latter the drone has to fly all the way back to the operator anytime a computation needs to be done making it a little less dynamic imo.

You need a proper use case analysis to do this with any hope of succeeding. You have to decide the goals of the system—that is, just what it is supposed to accomplish—before you can design it. Think of it as a “Why?” document. Why is this needed, how will it justify its own cost?

Then you need to figure out what it will look like to use it. A narrative of what it is like for any person that interacts with it, or the data it collects. You also need to check your assumptions now before you decide on parts. Just how much spatial resolution does the camera need to accomplish the analysis you expect this to do? What kind of lens will be required with what field of view?

You also need to create a power budget to match a practical power supply to the equipment you can expect it to power. Collecting approximate power requirements for each powered item and find out what you expect to need in terms of power.

You also need a weight budget. Do the same thing with all the hardware that will have to fly, this will interact with both the power budget and this one as the motor(s) keeping the drone aloft will have to lift everything, including themselves.

You’ll also need a link budget which will tell you have much power your transmitter will require, how big it will be, and what sort of antenna(s) you will need. This interacts with the other two, since the weight and power are linked to these choices.

And, this is glossing over a lot of things. This is one reason that in the real world engineers often use platforms that are already highly integrated with know capabilities. Reinventing something like a surveillance drone with the numbers you are throwing around is a project for a team of experienced engineers, not for a naïve student working alone—or even with a team of students.

I would suggest seriously scaling this back and maybe considering something like a commercial (consumer) quadcopter designed to assist road crews by mapping sections of a road as they do live pothole repair. This would provide more than enough challenge for one or more academic years.

 

You need a proper use case analysis to do this with any hope of succeeding. You have to decide the goals of the system—that is, just what it is supposed to accomplish—before you can design it. Think of it as a “Why?” document. Why is this needed, how will it justify its own cost?

Then you need to figure out what it will look like to use it. A narrative of what it is like for any person that interacts with it, or the data it collects. You also need to check your assumptions now before you decide on parts. Just how much spatial resolution does the camera need to accomplish the analysis you expect this to do? What kind of lens will be required with what field of view?

You also need to create a power budget to match a practical power supply to the equipment you can expect it to power. Collecting approximate power requirements for each powered item and find out what you expect to need in terms of power.

You also need a weight budget. Do the same thing with all the hardware that will have to fly, this will interact with both the power budget and this one as the motor(s) keeping the drone aloft will have to lift everything, including themselves.

You’ll also need a link budget which will tell you have much power your transmitter will require, how big it will be, and what sort of antenna(s) you will need. This interacts with the other two, since the weight and power are linked to these choices.

And, this is glossing over a lot of things. This is one reason that in the real world engineers often use platforms that are already highly integrated with know capabilities. Reinventing something like a surveillance drone with the numbers you are throwing around is a project for a team of experienced engineers, not for a naïve student working alone—or even with a team of students.

I would suggest seriously scaling this back and maybe considering something like a commercial (consumer) quadcopter designed to assist road crews by mapping sections of a road as they do live pothole repair. This would provide more than enough challenge for one or more academic years.

Thank you for this roadmap.

 

what about potholes in tunnels or on roads that have have structures built over them?

 

Consider first the camera resolution requirements, which will include resolution in the single centimeter range, at the distances stated. Certainly there are cameras and systems with that capability, and both NASA and the CIA have demonstrated them. Facial recognition at half a kilometer was the proof I saw.
BUT, given a guess at the SWP&C, (Size, Weight, Power, & Cost), such a camera package will be far out of sight, even for a wealthy municipality.

 

what about potholes in tunnels or on roads that have have structures built over them?

I am considering them as edge cases and not worrying about them for now

 

If you ask me, the city is being silly. Pissed off drivers (no shortage of those, I presume) and a widely advertised number (and web site) for reporting is a far simpler, less costly solution.

 

I'd be surprised if there isn't already an 'app' for reporting potholes?
Edit:
Oh, wait; there is.

 

I'd be surprised if there isn't already an 'app' for reporting potholes?
Edit:
Oh, wait; there is.

But the city would need to adopt it, no?