Dear fellows: I was talking with a friend of mine this weekend about an interesting application using IRIS+ (or a similar device) in order to inspect trains.

We are in Mexico and here trains are commonly used to transport raw materials like steel and coal from all over the country mainly to the USA border.

There is a big problem reported from years for all the companies doing this kind of services: robery. Thieves climb into the trains cars (train in motion) and extract all the goods in there so they would like to send a "drone" every hour or something like that in order to ispect train cars from above.

In France, they are already doing this kind of ispections, the main challenge is that train is in motion so I was wondering if you know some device in order to do this kind of application.

Thank you all!

Views: 620

Replies to This Discussion

Iris with a GoPro in the daytime or use a FLIR camera to detect heat signatures at day/night. You would need an FPV system and two people; a pilot and a watcher. 

Well, heck, I've done this myself here in the US.

If you had an operator on the actual train, they could release the multi-rotor while the train was in motion, inspect the length of the train, then return back to a landing spot on the train.  Sounds like a fun job!

Check out my video on this...  Now I did speed-up the video for purposes of YouTube, but the idea can be implemented.

If you want, I can post the raw video, it's about 8-10 minutes long.

Nah, one person is enough, just record high quality video 1080P or better, then review on board after inspection.  I did this flight from the ground, but had I been on the train, it would be the same idea.

Just hope you don't have a crash, cause the train probably can't stop for a mile or so, LOL.

Thank you for your video. This is pretty close to what I'm planning to do.

The main differente is that my inspection would be done with a pilot on the motion train, not out.

So the challenge is to take off the IRIS+ to inspect the car trains at a point A and to land the quadcopter on the coordenates where it was took off which are going to be point A' = A + velocity * time.

What do you think?

Yes, that would be very doable - take off from the train, then return to the take-off location (on the moving train).

Here's my thought (assuming you are taking off from the front of the train).  Take-off at the front, fly to a location perhaps 10m to the left or right side of the train.  Put the Iris+ in a PosHold Mode, and remain in that stationary position with the Iris+ pointed at the train.  The train will go by the Iris+ requiring very little pilot input (and this will provide the most stable video platform).  Then, once the rear of the train has passed by, fly the Iris+ to the front, then position it on the opposite side, wait for the train to pass, then fly back to the front and land.

Here's the tricky part, what to do in an RTL situation?  RTL won't work because your take-off position has moved.  If you know the train will be traveling at a very steady pace, you could have the final waypoint in an AUTO mission be somewhere different than where you took-off from.  The final waypoint could be programmed where the landing point "will be", then, instead of hitting RTL, put the Iris+ in AUTO, and have it fly to the final, pre-determined location.

This seems like a fun project!  Good luck.

how fast will the train be moving? 

How long is the train? 

The problem might be fight time if you are hovering in one spot then trying catch up to the front of a moving train. 

If the train speed was up over 45mph, with a head-wind, could definitely be a problem, getting back to the front.

My F450 can easily do 50mph carrying a GoPro H3 and H3-3D gimbal (no wind), I suspect the Iris+ might have similar speed capabilities???

Flight time shouldn't be a problem.  A 1mile long train will pass a stationary point in 2 minutes traveling at 30MPH, the Iris+ traveling at a return (ground) speed of 45MPH, you should be able to reach the front of the train in about 4 minutes (there's a bit of algebra at work here, because the train is moving forward and the quad is moving forward too but at a relatively slower speed compared to the train (about +15MPH), so it will take much longer to return to the front.

Here's my algebra, let's see if I'm right?:

Train @ 30MPH = 44ft/s

Quad @ 45MPH = 66ft/s

equation:  (44ft/s) x t + 5280ft = (66ft/s) x t

Solving for t, t = 240sec, divide by 60 sec, gives 4 minutes for the return.

I think its doable, 6 minutes per side.

Now if train is moving at 40MPH (58.7ft/s):

The stationary pass will be 5280ft / 58.7ft/s, or about 90secs

the return equation:  (58.7ft/s) x t + 5280ft = (66ft/s) x t, or about 720secs

Total Time 90s + 720s, or about 13.5 minutes, per side.

Also, make sure your GEO_FENCE is turned off, otherwise the Pixhawk Flight Controller with go into RTL if you breach the fence distance... assuming you have this setting turned on as a FAILSAFE.  In the 30MPH train example above, your total flight distance will be almost 16,000ft.

You might want to get a wind speed gauge (Anemometer) to determine actual head-wind speed (cumulative between the train and wind).  I would put this on a stick far enough out so as to not measure the turbulence air near the train.

You could add this piece of information in your algebra equation to account for any head-wind (tail wind will be to your advantage).

You would definitely need to disable nearly every failsafe since they're all based on the drone environment being stationary.  If you have a control failure, it's basically has to crash.  No return to launch. No geofence.  You'll need to accept the financial risk.

You'll need to consider the range of your transmitter too. You need to be able to control the UAV for the entire length of the train plus some margin of error.  A mile or two is well outside the standard range of most stock control systems.

On the same token, your FPV system will need the same range.

@Pedals - Definitely good point on the transmitter / FPV range capabilities.  My Futaba T14SG can reach out about to 5000' to 8000' max, so I was thinking a mile long train was doable.

If you defeat the transmitter failsafe and one were to use AUTO mode for most of process, you could have the quad wait for X amount of time at the "PosHold" monitoring point (location you fly to the side of the train for the starting point), then after the waiting period (manually calculated by taking train length divided by speed), the quad would then fly to a location where the front of the train will "be at" for the return (you could enter this location prior to taking off using DroidPlanner and a cached Map).

I don't know if the Iris+ has RSSI monitoring, but if it does, you could add that to your OSD to monitor status of your receiver's signal strength. 

I think this would be a fun project.  It is similar in nature to a distance test I was going to run on my hex.  My hex can fly about 36 - 40 minutes on a 16,000mAh 6S battery averaging about 17A at 30MPH.  My hex has a Pix F/C and I was going to see if it could fly 10 miles each way based on an AUTO program that was pre-loaded in the F/C.  I found a location where I could do this with only a 50' elevation change.  My plan was to just drive beside the hex while it was in flight (just to monitor it).  Once it reached the 10 mile target waypoint, it would turn around a head back to the starting point.  Not as complicated as the train task, but similar in determining all the parameters it takes to make it work.  The challenge is what is so cool.

Si estas pensando de usar la sistema de APM para su proyecto, tienes que tener
en cuenta que el software que maneja los vehículos tiene una serie de cosas que
revisa antes de armar para el vuelo.

Uno de esos es revisando si la velocidad esta mas alta que 50cm/s. 
Busque "bad velocity" en este articulo pa entender.

Esto no quiere decir que es impossible, si no que vas a tener que modificar un
parte critical del software que maneja la sistema operativa del vehículo.

Me parece que seria mas util usar una sistem para detectar un cambio electro/magnetico
cuando se introduce un nueva carga al tren. Usando sensores en cada caro, hasta podrías
identificar en que caro se esta detectado el cambio.


© 2019   Created by Chris Anderson.   Powered by

Badges  |  Report an Issue  |  Terms of Service