To borrow a title from a former chancellor of our alma mater; in this post we'll provide a paraphrased history of the project from inception to the present day.
Intro and early days
Getting straight into it - the idea for the project came from a conversation between Tom M & Tom A at a friends' wedding in 2014 - thanks Andy and Michelle!
During the project we've been very fortunate to be invited to the Natural History Museum to give a talk and have a stand at a couple of events. One of the questions that comes up most often is "what company or university are you at?" We're not. The project is simply a very enjoyable hobby. We want to help further bat conservation it's a great way that we can assist with this using our respective skills. Tom A is our bat expert (and is now a qualified and competent drone pilot) and Tom M is our drone designer, builder and operator.
We think this goes to show that if people are willing, you don't need the support of a big research grant or institution to push boundaries. Our advice is to "give it a go and see where it takes you".
We knew that quadcopters produced a lot of sound but didn't know how much of this was ultrasound. Our first tests showed us that it was a considerable amount. This ultrasonic interference has been a running theme through the project. The problem lies in the signal to noise ratio of the bat call. If there is too much noise from the drone we won't be able to 'hear' the bat call.
This meant that we had to separate the drone and the detector recording the bat calls by about five meters. The highly technical application of a piece of string was employed though this wasn't without it's problems as you can see in the video below.
The problem of the oscillation arose due to the high weight of the detector. We were glad we used a water bottle of the same weight as the detector in testing! The detector and recorder weighed about a third of the quadcopter mass. This led to loss of control and meant and it wasn't an option in this configuration.
In going back to the drawing board we wanted something that produced less noise and could comfortably fly with the payload of the detector. The obvious choice for us was to move to a plane; it had the added bonus of being able to fly for longer too (about 8 mins for the quad and 25 for the plane).
Last time we picked up the Batboat we put in an autopilot and managed an autonomous run (check out the last blog post here for more info).
We now wanted to move on to fine tuning the autopilot and some real world field testing.
Tom A had recently performed a survey along a transect route that included a decent stretch of the River Thames in Oxfordshire. A number of species (Daubenton's Pipistrelles and Noctules) were picked up during the survey so we knew that this was a good stretch of river to test the Batboat on.
What are we looking to test?
To this end we headed down to the river during the day to assess the autonomous navigation of the boat along the river.
We set a waypoint route for the boat that headed up river, turned around, then came back down river past the starting point and then another turn and back to us. All was well on the up river stretch and we were able to walk along and see the boat from the river bank. At the turning point, the boat struggled to make a small enough turning circle and ran into the near bank. Fortunately, we did plan for this and had brought some swimming things so we could fetch the boat if necessary. A short cold plunge later and we were able to run another test.
We're not experts on the navigational algorithms that the code contains (on the autopilot) and although we went through a number of variations with settings we only managed to make a couple of successful runs up and down the river. We think that a directional propulsion system would have more luck in steering the boat successfully; this is something we'll investigate in the future.
Despite this, we knew that the boat could at least navigate either an up river section or down river section successfully so we decided to head back out during the evening for a field test.
Lights were fitted (red for port, green for starboard) and we had no issues seeing the boat in the dark. First hurdle overcome.
We'd set up a long run (circa 400m) up the river and then back to us so we could follow the boat from the river bank. We really hoped that the boat would make the turn on this one as it was now dark and neither of us fancied another trip into the water to fish out the boat!
With bat detectors in hand beeping away, we knew that we had a good chance of success so launched the boat and flipped the switch for the autonomous navigation to take over. Leaving all of our other equipment by the river bank we excitedly walked a parallel path to the boat and kept our fingers crossed that all was going to be well. A few minutes later and we'd reached the turning point (fingers now extra crossed). The boat started to make the turn away from us and toward the far bank. All was going well until we realised that there were some low trees with branches in the water which were lit up by the lights and the boat was now powering into them. We flipped the autopilot off and back to manual control but it was no good, we'd managed to get the boat stuck on the far side of the river.
Somewhat dismayed at what we knew was coming up we trudged the 400m back to our equipment and picked up the wetsuit. The bank was fairly steep opposite the boat so it took a little hanging onto trees and squelching through river mud to reach some quite chilly open water. With the head torch on low to avoid too many insects being attracted I swam over and pulled the boat from the branches. Boat successfully recovered? Check. Swam in the Thames at night? Check.
The question that seems to come to mind is: was it worth it?
To answer that, have a look at the selection of recordings from the evening run below.
In relation to our aims for this experiment:
The autonomous navigation of the Batboat - is it able to successfully navigate a loop on the river?
Not reliably. We've seen that it is possible though requires refinement of the navigational algorithms that control the turn or a different physical set up of the boat.
Peersonic have kindly provided us with a new microphone which has a filter on that suppresses frequencies below about 20 kHz. We're hoping that this means we have a better signal to noise ratio given that the motor/prop puts out noise at (generally) below 20 kHz.
Yes. It appears that the new microphone is suppressing the noise from the motor and as a result there is an improved signal to noise ratio for the bat calls.
Successful use of the Batboat at night - are the lights ok? Can we see the boat?Yes. No problem seeing the boat from the bank or at approximately 100m away.
Recordings of bats - can we get any, are they useful (ie good signal to noise etc)?
Yes. Very happy with the signal to noise ratio of the calls, there's is noise showing in the spectrograms and importantly it doesn't interfere with the calls which are very clear. Worth noting as well that we didn't see any change in the bats flights when the boat was near them. They appear to ignore it.
In conclusion, we're happy with the results from this set of testing. From a search we can't find any instance of anyone attempting to record bats from an autonomous boat so we're claiming a world first on this one.
We'll be working on this system to improve reliability and then move onto further field testing.
Keep your eyes on the blog for more updates!
We have made a lot of advances on the detector side of things so decided to re-visit the use of the quad-copter, something we had moved away from due to issues with motor noise and the weight of the detector.
We suspended the Peersonic detector under the quad and flew a small route of approximately 200 meters along tree-lines surrounding a field. In the corner of the field we picked up bats. This is the first time we have picked up high quality recordings of bats using our setup!
The flight path of the quad-copter
Sonogram of our first recordings
Video of the flight with audio from a Bat Box Duet (held at ground level along with the camera)
This is a real break though as the noise we see from the propellers is clearly significantly quieter that the calls we recorded. To reduce this further we hung the detector a little lower (initially 2m and changed to 4m) from the quad and got some even better calls from Common Pipistrelles (Pipistrellus pipistrellus). We know these are common pipistrelles because of their 'hockey stick' shape and because the peak frequency is at around 45kHz.
We are going to continue to explore the potential of both the quad and the fixed wing as platforms for bat detectors.
Watch this space!
After our success with the Peersonic detector working in-flight (see our blog post here) we decided to push on and get some actual field testing done! This will be the first time we've attempted to use the UAV to record actual BATS!
The first thing to do was identify an area in which is likely to have some bats flying around and that we can fly the UAV at. Fortunately, our flying site in Oxfordshire has a lovely line of trees with woods and hedges at both ends which look quite promising.
During the day we set up a flight path which took us down the tree line, over a hedge at the end, looped round the wood then cut in towards the tree line again heading back towards us and past the wood at the end with a final turn taking us over a hedge and back to the starting point. The whole route took about 6 minutes 30 seconds to fly.
Although the description is all very well and good a picture of the route might be more useful:
We flew this a number of times to ensure we weren't too close to the trees or in danger at any point. To do this we used a FPV (first person view) set up on the plane in order to assist in distance perception to trees. It turned out that this was fairly crucial as during the first run we were a little close to the trees. After some minor changes to the flight path we repeated the route and were happy with it. We decided on an altitude of 15 m to make sure we'd have more than enough clearance for any objects. Here's the plane's view of the flight path:
On to the night flying!
With the route set, we were ready for the night flying with the detector. A bit of fiddling later and we'd set off on our first pass of the route. It's a strange feeling watching the lights of the plane sail off into the distance. It's great to see how stable the plane flies at night as well when the wind was a little lower.
After two passes we decided to take a walk up to the woods at the start of the tree line to see if we could find any bats using our other detector. Happily we noticed one or two commuting down a hedge line that connected to the woods.
A short re-route of the flight plan later and we had a flight running down the hedge line at an altitude of 8 m and 15 m horizontally from the hedge.
Have we recorded a bat?!
When looking through the sonogram, we noticed this:
Between 204300 and 204400 ms there's what looks like a 45 kHz bat call. The sound lasts for a duration of approx 7 ms and has a classic hockey stick appearance. It's also at bang on 45 kHz which ties in with a 45 pip.
There is however, a lot of ultrasonic interference from the motor/propeller and we want to be sure that this is a bat.
HERE'S THE LINK TO THE RECORDING:
If you have any experience we'd really appreciate your thoughts. The full .wav file is accessible via this link (right click - save as). Please feel free to download let us know what you think.
As always, we'd like to acknowledge Peter at Peersonic.co.uk for his constant support and providing the recording equipment that has been so useful in progressing this project.
Thanks, Tom & Tom