I would like to show what we have done here in Namibia in developing Inexpensive UAV systems which we are beginning to actively use in conjunction with Namibian Ministry of Environment and Tourism, tasked with wildlife conservation in Namibia, and also for photography missions for the local quarry mines and Salt mines.
Shown , in order, are three aircraft, GUPPY, fully designed from the ground up and built by me, HORNBILL, built from a semi built kit, the Lanyu-100, and the third, KIWIT, an electric flyer of unconventional wing design.
HORNBILL is fitted with an RF tag tracking system and antenna and flew many autonomous, beyond visual range flights tracking tagged animals, especially Rhino as an aid to the National Parks anti-poaching campaign. The next trial will be with an Infra-Red camera fitted to aid the detection of poachers at night.
Hornbill basic Specifications.
- Wing Span - 2.4meter
- Engine 55cc Twin Boxer.
- Weight 15kg with 3liter gas
- Max TOW - 17Kg
- Endurance 1H 40minute with 0.5 liter spare @ 24m/s
- Min safe speed 14m/s @ 16deg AoA.
- Stall speed 10m/s @ 25deg AoA.
- Max flight speed - 34m/s
it was a lot of fun! See the report: waterberg_report.pdf
GUPPY was designed from the ground up, a full 3D model developed on CAD, and designed to be made from good aircraft grade ply. However, it was built with plain Marine ply in the end, as and exercise in cost reduction. It was designed to carry a hi resolution camera , which was subsequently flown to create geo-referenced imagery for the Quarry miners, as well as for the Salt Works. GUPPY has tremendous STOL performance and can land/takeoff anywhere.
GUPPY Basic Specifications:
- Wing Span - 2.8meter Clark Y 14%
- Wing Chord 600mm +80mm for Junker Flaperon
- 2 Flaperon per wing, each with own servo
- Horizontal stabiliser - 450mm each half span Clark Y 12%
- Horizontal Stab Chord - 220mm
- Each stab half fitted with own servo
- Engine 100cc Twin Boxer. (3W)
- Weight 26kg empty
- Max TOW - 42Kg
- Max Fuel - 8.5liter
- Endurance 5hours with 0.5 liter spare @ 35kg @ 24m/s
- Min safe speed 12m/s @ 20deg AoA.
- Stall speed 10m/s @ 25deg AoA.
- Max flight speed 42 m/s
The electric flyer was actually two Aircraft, PiPiStrello, and KIWIT. Both share wing profile type, Jedelski, and both are V tail. Both are made fully from composite materials.
PiPiStrello has straight wings with dihedral, and was fitted with ailerons, not normal on Jedelski wings – resulting in excessive adverse yaw so was canned…
KIWIT and its prototype flew many flights tool to verify the existence of a Vulture Chick in a nest in a tree out in the desert. We annually undertake a conservation mission where we currently fly in a Cessna 180, spotting Vulture nests in the Sparse Namibian desert regions. When we spot a nest, we fly closer to see if there is a chick, and if so we GPS mark the spot. Often the tree is among a dozen others, so the exact spot is 'close by'..but not easily found when we return in ground vehicles to capture, ring and tag the chick. So we used KIWIT to overfly the nearby area and find the chick again.
See the article in Raptor News - (page 2) Raptor_News.pdf
We have also perfected fully autonomous land and takeoff with KIWIT.
Have a look at these videos for the Auto Launch and land:
Kiwit Basic Specifications:
- Wing Span : 2.4meters
- Weight excluding payload - 3.5Kg
- Max payload capacity - 1Kg
- Endurance - 1hour
- Cruise speed - 14m/s
- Max speed - 20m/s
- Stall speed - 6 m/s
- Takeof in 3meters in still air.
- All composite construction, cabon fibre and glass
And these videos for the flights of GUPPY and HORNBILL:
Have a look at the postings under Aircraft Platforms and Autopilot Hardware for more info.
The autopilot fitted to all these aircraft is called the 'NamPilot' and is based on multiple modules, namely:
AUTUPILOT: an AVR processor based module, 8 PWM inputs from RCS, 7 PWM outputs to servos. This module does the full flight stabilization and navigation/guidance for auto flight, and stabilization for manual flight with autostabilization - mostly used when tuning the control loops.
IMU - Also AVR based, with 3 axis Gyro's, accelerometers and a Bosch I2C pressure sensor.
AIRDATA UNIT: Also AVR based, a smaller processor, with another Bosch I2C pressure sensor, and an analogue static pressure sensor ( Freescale) and a Dynamic pressure sensor for airspeed. (Freescale)
MAGNETOMETER: Also a smaller AVR processor, with a Honewell 3 axis magnetomet, configured in a balancing bridge circuit, ie, the null strap in the device is used to balance out the ambient magnetic field in a control loop. This gives VERY accurate performance, with almost no temperature drift at all, the bain of magnetometer performance!
The Ground Control Station - This is a PC based system and can be anything from a simple, reduced functionality Laptop based system, to the one depicted in the PDF files listed below. This system interfaces to a datalink antenna, with optional autotracker for the antenna, tracking theUAV in flight. It also interfaces to a dual video receiver, built into the video RX antenna. The video from the UAV camera is available in real time on the GCS, and can be recorded. A dedicated configurable Payload control panel is on the GPS operator position, allowing control of the UAV stablised camera, steering, setup, etc. All flight plans are prepared on the GCS, with a map of the flight regime, and all flight paths are shown, as well as the track flown. Numerous flight patterns can be selected, figure 8, square, straight line, circle, loiter, point to point flight plan, etc. Aircraft control loop tuning is also done in real time, during flight, from the GCS.
Take a look at these PDF’s for more info:
imu_flyer_full.pdf autop_flyer_full.pdf gcs_flyer_full.pdf
This is far too much fun for just two people!!
we are in Swakopmund, so you were 50km from me! You would be most welcome!
I will look at the videos too..
Hi again Joe
Fascinating stories - I'm very happy that someone is applying this technology to securing the rhinos. From what I've heard, as security tightens on South Africa, more and more of Namibia's stock will be targeted - seems like you're getting ready for it.
I flew near Wilhemstal (just out side Karibib) on a 9000ha farm. I just had my flying wing there, flying FPV. Here's some footage. I've also been flying in the dunes and at the coast. Here are some clips: https://www.youtube.com/watch?v=cE5tkZsreho
I'd love to come out and visit you! Where are you based? We go up yearly, sometimes more, so I'd go out of my way to pop in.
Thanks for your kind words...Yes, we are quite deeply involved here with the Ministry, esp on the security and anti-poaching side. We are going to be putting in a large trial system in two of the big parks here, over the next 5 months. The trial is for 15 months, and is being sponsored by WWF and Google...It includes the deployment of up to 6 UAV systems, with still cameras, Infra-red video camera, etc, as well as training of all the staff, systems operators, observers and so on. It also involves the generation of processes by which we hope to determine the more successful monitoring methods for game tracking and poacher detection, so I am very excited about it! So far we are achieving 20km range with our datalink and video systems, with a tracking ground antenna ( long helical) so that should be a good start. We will see...!!
We have tried game counting on a number of farms - I am afraid it does not work - we mostly succeed in upsetting the ranch owner because his statistics are completely thrown out after our flight counts! Normally it is the Ranch owner who wants it done , because he believes that he can now do it every 3 months, and at hugely reduced cost, and its an exciting toy, so he wants it shown and then he will buy...And then he finds his game is reduced by 60%!
He starts blaming the system, the camera, etc and gets very irate - his usual chopper pilot just sits there and smirks - thinking to himself he must now push up his hourly rates....
We did a count where the chopper followed the UAV, trying to stay within 150 to 400meters from it , chasing out the animals - that worked MUCH better, but there is another problem - the animals are in herds, so when you scare out a herd, you need to follow the herd. If you are under autopilot flight control and flying transects, by the time you have flown the current leg and return to the area on the next leg, the herd has left the scene, or run into your new photo leg so you add the same counted animals to the new photo....Does Not Work!
We have implemented flight modes ( via autopilot and GCS) that you would not believe - When we spot animals in flight, we can initiate all sorts of loiter patterns, figure eights, ovals, straights, zig-zags, etc, etc...to try and capture and cover the herd. We are as often as not, going in the wrong direction, going to slow, or to fast, etc. We can also fly in 'stabilize' mode, where the autopilot just keeps the plane safe, keeps to airspeed and altitude, and we steer with a joystick. It still does not work - we can easily put the plane where we want like this, but that is the problem - where do we want to put it? The camera field of view is far to narrow , and the resolution at 100m height , with the plane bobbing about, and the animals running - you have no idea where to place the plane!
What this method does do very well indeed is make me become violently 'airsick'! You become so absorbed in the video and focus so intently on the rapidly gyrating ground view that you actually lose connection with the ground, and at a point..there it comes!
What a game!
Where were you flying here in Nam? Why did you not come visit us on this West Coast Pimple??? Always trying to bribe people to come visit - they get free board and lodging in exchange for bringing me a list of goodies ....Only thing the hardware stores around here sell is cement and screws that come pre-rusted...
Thanks for bringing the hard realities home. I took my small FPV plane to Namibia over the holidays and did some flying on some game farms. My brother in law was hoping I could help him spot game to hunt and count. However it was tough going, I could not spot anything in real time, even after flying looking at the HD - I could not find a single buck even though there must've been many. It's true what you say - when you're in a chopper counting game, it's the movement that catches your eye, so perhaps a UAV will be hard pressed to beat a manned chopper.
I've been asked to bring my Skywalker with Arudplane on in April to do some surveys, but I'm starting to have doubts.. despite the original blind optimism it might be a waste of time trying to use it as game counting.
All the best, love you work.
No Problem Marc.
Forgive all my typo's - fingers overtake the brain at times...
If I can help or give any info, please shout - My experiences and suggested approach is/are not the only ones that work; I am sure there are many folk doing similar work in other ways , and keep very quiet..
Gisela and Joe Noci,
NamPilot is a commercial product? Where I can find more information, prices, features etc?
Thank you, Joe. That provided me a good context to approach my project idea from. Your "ramblings" are more cogent and useful than many others' carefully prepared explanations, I think.
Those are difficult questions!
There are a number of inter-dependent requirements that one has first has to consider and decide on certain trade-offs.
First off, what sort of ground resolution do you require? If you want 2cm GSD, you have to fly very low, the camera ground footprint is small, and you have many legs to fly to cover a given area. At 2cm we would takeoff, fly a 1kmX1km area, and land, within 30minutes, and have a good few hundred photos to cope with.
At 15cm GSD we would cover around 4 to 5 times the area, in the same time, with a similar qty of photos. I say 4 to 5, because the maths gives 5 times, but you use huge amounts of battery to climb, and even though you have to descend again at the end, its to late, running out of battery at 5 times the area, at 550meters up, is not on.
The shape of the area also has an effect on time - long legs with few turns at the end are best - our flight planning software allows the generation of many different flight patterns so we can fly long, or short legs, etc. The flight planning software really needs to be very versatile to allow the user to manipulate the flight as much as possible - for example, if the are is triangular, you probably want to fly paralleled to one of the longest legs, but as you fly and approach the opposite leg, it keeps shortening, so you want to be able to fly a continually shortening leg on that side, e, not simple a rectangle that encompasses the triangle... That does mean that you have to fly very sharp 'dog-legs' at the turnaround point, and need to extend that leg out from the end, else you are not straight and level when approaching the return photoleg...that sometimes means that you do not save any time or battery, since each end is extended, and very sharp turns use lots of battery - a turn is a loss of lift, requiring a pitch increase to not lose height, and a throttle increase to not lose speed due to the pitch increase, etc, etc. So, really, the planning software must give you all the options, and allow you to swap between various scenarios to decide for yourself what appears the most efficient and stable flight plan.
There are other factors to - some planes claim a 60 minute or 90 minute flight time. Our SurVoyeur flies 30minutes at 400m ASL, and 28deg C. Does that mean My system is no good?. Not really - We fly at 72Kmh ( 20m/s) Most of the 60minute flight systems cruise at 12m/s to 15m/s.
We would cover approx 35km distance in 30minute. A 60minute flight at 12m/s covers 43km. Sure is covers more, but at what cost? the slower the flight speed the more susceptible to slight updraft, thermals and gusts. That means the camera is looking all over the place. So one really needs to look very deeplay at the system specs before we can say this is better than that...I have flown many thousand of hectare photo flights now, and we have tried many variations. We have flown at 3Kg, from 14ms/ up to 24m/s, all the way up to 5kg at the same speeds - the optimum was around 4.2 to 4.4kg, and 20m/s - that combination with our autopilot tuning gives the best results.
As to what the optimal area size is for a small UAV, there really is no single answer, I am afraid..
If you need 2cm GSD, then 100hectare is a good size, with a flight time of 30minutes.
If 6cm GSD is ok, then the same system would be good for 300 to 400hectare per flight.
This is one of the main reasons that I always tell people wanting to use such a system as a substitute for manned fixed wing/chopper wildlife counts that they will not succeed. To detect a buck standing 1meter tall at the shoulder, in its natural colour habitat, you will not get away with lower resolution than maybe 3GSD ( if its not under a tree or sidling a bush) That means MANY photos of the area. Manually scanning the photos afterwards id very tedious and trials have shown that a trained eye will lose up to 40% of the count! The same eye, repeating the count with the photos randomized, has a count that varies from 25% lost to 58% lost , over 9 repeated counts!
The main reason is that the photos are inanimate and flat. It is very difficult to spot animals in their natural habitat - you can easily spot a black elephant walking over white desert sands, but let that elephant bath and cover hinself with the natural colour dust, and you will not see him! And another very important reason is that in the photo the animal does not move. It is the slightest movement that catches the eye, esp the corner of the eye, and the brain rapidly processes that to steer the eyes on target. In addition, the eye/brain connection allows extremely wide angle field of view, with vere rapid zeroing in and focusing on areas of interest. None of that can take place with dead photos....
I have rambled on enough...
Joe - this is really great work you've done, and thanks for sharing all the details you have. In the missions that are done to monitor animal populations, what are the practical limitations in terms of total area that can be surveyed with the types of drones you have produced? Based on your spreadsheet, the total number of images climbs crazily as you go from tens of hectares to hundreds or thousands of hectares (scales at which wildland grazing animals might need to be observed). In your experience, how large of an area is really suitable for these light drone platforms?
have you considered straight tube fuselages? excellent work a very good example of UAV technology in use. Well done.