Autonomous precision landing to a charging station?

[mikcarlsson]

Hello,

I am thinking of different possibilities of precision autonomous landing. My goal is to have accuracy of approx. sq. with 0,5 m (2 feet) side. The idea should be taking the bird with GPS above a charging station (exists today) and let it make a secure landing. The problem is as you know the GPS precision of approx. +/- 3 meters (10 feet). Do you have any suggestions or know about an existing solution?

Kind regards,
Mikael

 

[LandYachtMedia]

The way you would do it with the existing hardware is to create a differential GPS beacon at the landing site that can correct for the GPS errors and bring your positional accuracy down to millimeters. To do this you would need to be able to spoof one of the WAAS satellite IDs. Those solutions do exist. You'll need a few thousand dollars to build the setup.

Anything else would have to be put on the craft. Whether it be LIDAR or vision based position augmentation, it will be expensive and cost you in additional payload weight.

 

[locoworks]

what about using a visual system once you get close to home?? the inspire and others use a visual type reference to hold position indoors, maybe a way to make a second camera look for a specific image to orientate itself above/over???

 

[MacCool]

Easier to use video recognition IMHO. Camera pointed down (or dedicated downward-facing camera) connected to some sort of phase-contrast distinguishing software. GPS will get the Phantom within range of the landing target, which would have some kind of optimally-recognizable pattern printed on it (or LEDs or something like that) that the system would recognize and track to landing. I just don't see how you'd feasibly get GPS to give you the accuracy necessary to get 1-foot plus/minus accuracy. It's going to have to be optical.

 

[jadebox]

Easier to use video recognition IMHO. Camera pointed down (or dedicated downward-facing camera) connected to some sort of phase-contrast distinguishing software. GPS will get the Phantom within range of the landing target, which would have some kind of optimally-recognizable pattern printed on it (or LEDs or something like that) that the system would recognize and track to landing. I just don't see how you'd feasibly get GPS to give you the accuracy necessary to get 1-foot plus/minus accuracy. It's going to have to be optical.

A friend and I were brainstorming this one day. I started to say "coworker," but we wouldn't waste time at the office talking about such things.

We were thinking about using a quadcopter as a part of a home surveillance system. It could rest in a closed pod on the roof then pop up to scout around the yard when commanded. It would return to base to recharge and for protection from the elements. (If anyone makes a fortune from this idea, PM me for my PayPal address so you can send us our cut.)

We decided on your solution as the most practical. Some multi-rotor aircraft already use a similar method to help them hover without drifting.

Other ideas included a big funnel to guide the 'copter in. But, that wouldn't look good on your roof. A smaller one, though, might work in conjunction with the optical solution.

-- Roger

 

[kitari]

Are you talking about something like an inductive charging station for the phantom? I think someone was talking about something like that a while back, only they were talking about having a group of people that could charge it for you instead. I think it would be a great idea if the drone delivery service actually took off. You could put charging pads on light poles and make them like drone refuel points. Sure it'll cost money to keep all those drones charged, but by then, drones will probably have an identity rfid chip that can charge your account by the kilowatt/hour or something.

 

[mikcarlsson]

Thank you for your suggestions! I believe that computer vision is the way I will work on. I have thinking of using induction but that will put weight on the machine and is not as cost effective. Using the camera and guide it safely to the "drone port" needs some programming. I'll keep you updated.

Mikael