Hello,
I am part of a university design and build team who have been tasked with building a fully autonomous aerial vehicle capable of delivering a 2kg payload from A to B over 2km and also must be capable of flying a figure of 8, all without manual control. I am asking for help from whoever may be reading this on how we can do this, such as what sensors we need, what code we need and what board we need. We are probably going to go with arduino equipment and we have experience with C, C++ and python. We have decided to build a quadcopter, it must take off, perform a figure of 8 and then travel 2km in under 2 minutes and drop a 2kg payload (bag of flour) on a target which is an 8mx8m green square with a red A4 piece of paper in the center with an alpha numeric code printed on it. The largest challenge we are currently facing is how do we get the quadcopter to "see" and recognise the marker and therefore trigger the descent of the quad and the release of the payload. Our budget for "off the shelf" components is £1000.
Any and all help/advice is greatly appreciated, thank you.
Also, in case your interested, this is for a national inter-university competition run by the institute for mechanical engineers (iMechE).
Because the task involves computer vision I would recommend NOT using a basic Arduino. I would use a Raspberry Pi (or similar small LINUX board) with a camera.
You will need GPS for rough location. I'd use a 9DOF sensor board (accelerometer, rate gyroscope, magnetometer) for balance and direction. I'd use sonar or IR distance sensors for fine altitude near the ground.
A 2kg (4.4 pound) payload is pretty hefty. With flight batteries, motors, frame, controller board, ESCs, sensors and navigation this is going to have to be BIG quadcopter. You might consider building a hexacopter (6 motors/props) or an octocopter (8 motors/props) to better carry your load.
There is a wealth of open-source software/information/products here. I have scratch-built and am currently flying a wood-framed quadcopter that uses an arduino Mega 2560 R3 board and an Aeroquad flight board (accelerometers/gyros/magnetometer/barometer). All software is open-source and readily available. They are also making progress toward autonomous GPS navigation where you could enter "way-points" into your multi-rotor and have your vehicle fly the course you have pre-loaded into it.
Also... do you literally have to get the vehicle to "see" the payload drop-off grid, or could you use GPS to get you enough position accuracy? Image processing for recognizing and navigating to the red paper on the green square could be quite software and processing power intensive.
Good luck your project and keep us posted on your progress. Many of us here would enjoy hearing/seeing details of your efforts.
It's something I've often pondered. I think this is quite doable with a gps module. I'd be inclined to go with a derigable balloon. At least then the basic stability of keeping it in the air should be fairly simple.
KenF:
It's something I've often pondered. I think this is quite doable with a gps module. I'd be inclined to go with a derigable balloon. At least then the basic stability of keeping it in the air should be fairly simple.
Given the payload and time requirement, I'd be inclined to look at using a large RC style model helicopter. You can use a glow fuel engine to avoid battery issues and control the aircraft with ardupilot which would give you range and gross gps navigation. Then you're left with solving the computer vision problem with a Pi or the like for accuracy at the target.