Close

Hack Chat Transcript, Part 1

A event log for Aerial Robotics Hack Chat

Into the third dimension

dan-maloneyDan Maloney 06/08/2022 at 20:100 Comments

Dan Maloney12:00 PM
Welcome to the Hack Chat everyone, let's get started. I'm Dan, I'll be modding today along with Dusan (I think) as we welcome Nick Rehm to talk about Aerial Robotics.

Dusan Petrovic12:00 PM
Hi Dan!

Nick Rehm12:00 PM
Hey everybody, can you hear me?

Dusan Petrovic12:00 PM
Welcome everyone!

Joe Stavitsky12:00 PM
No but we can see you :)

Boian Mitov12:00 PM
Hi @Dan Maloney and @Nick Rehm :-)

Dan Maloney12:00 PM
Hi Nick, thanks for coming on today. can you start us off with a little about yourself?

pascal.fust12:00 PM
nothing to hear, but I can read you

Johan von Konow joined  the room.12:01 PM

Dusan Petrovic12:01 PM
+1

Nick Rehm12:01 PM
Sure Dan

garrettmagma joined  the room.12:01 PM

Nick Rehm12:02 PM
I've been into radio control aircraft for about 14 years, so it was really fun to see the drone revolution happen in front of my eyes growing up. Graduated from University of Maryland in 2019 with BS in aero engineering, then went on to masters of aero/rotorcraft with a focus in aerial robotics/dynamics/control

Nick Rehm12:02 PM
My favorite era of aerial robots / done was the multi-wii flight controller days where I had to convince my mom that I needed to tear apart the wii controller to build a drone

Joe Stavitsky12:03 PM
Heh, I screwed around with that briefly

Dan Maloney12:03 PM
Did she buy it?

Nick Rehm12:04 PM
It finally clicked for her when she saw a wooden tricopter floating around the front yard haha

Joe Stavitsky12:04 PM
Lulz

Dan Maloney12:04 PM
Better to ask forgiveness than seek permission, right?

anfractuosity12:04 PM
was the wii controller used to control the drone, or in the drone out of interest

Myself12:04 PM
With an accompanying triumphant scream of "IT'S ALIIIiiiiiive!", I presume?

Nick Rehm12:05 PM
I think my parents loved the idea of RC because it kept me both in my room and away from video games

RichardCollins12:05 PM
@Nick Rehm Was your masters at University of Maryland? (College Park years were some of the best)

Nick Rehm12:05 PM
We used the IMU from the wii motion plus back before IMUs were cheap as heck

anfractuosity12:05 PM
cool :)

pascal.fust12:06 PM
nice hack

Joe Stavitsky12:06 PM
@Nick Rehm do you mind saying where you work?

Nick Rehm12:06 PM
Ahh yes, masters from maryland as well--I went straight from BS to my MS program because I was already well acquainted with the rotorcraft center team

Nick Rehm12:07 PM
I'm at JHU APL working as an aero engineer

Nick Rehm12:07 PM
I'll leave it at that, but obviously there is some overlap in hobby/job skills

Joe Stavitsky12:07 PM
@Nick Rehm awesome

Cliff W joined  the room.12:07 PM

Sheridan-Tech12:08 PM
Could you tell us which kind of projects you have worked on

Sheridan-Tech12:08 PM
?

pascal.fust12:08 PM
@Nick Rehm What do you think about the optimization of aircrafts for VTOL/fixed wing UAV ?

Nick Rehm12:09 PM
Sure, one project I have out there is an open-source flight controller which I use on every single flying thing I make: https://github.com/nickrehm/dRehmFlight

pascal.fust12:09 PM
good stuff

Nick Rehm12:09 PM
It is meant as a teaching tool for people who aren't too familiar with software dev

Yohan Hadji12:10 PM
Hey Nick, what is in your opinion the next big challenge in the domain of small autonomous flying robots?

Nick Rehm12:12 PM
@pascal.fust I think optimization is extremely hard until the system constraints are defined really really well. For example, "we need a vtol that can fly for a long time".... Can that vtol transition and fly on a wing? Are there loitering constraints? Funny enough most problems not involving the need for a loitering camera can be solved with a regular fixed wing

Joe Stavitsky12:13 PM
@Nick Rehm I mean, in that formulation if you can circle small enough you can call it "loitering" I think?

Nick Rehm12:13 PM
@Yohan Hadji I think the big challenge always has been and will continue to be onboard state estimation. It's getting better with better sensors and lighter computing, but still not at the level of motion cap facilities

pascal.fust12:13 PM
I agree. But, I can also observe that most VTOL aircrafts out there still don't benefit of the fact, that takeoff/landing low speed lift capabilities are not required

Joe Stavitsky12:14 PM
@Nick Rehm sorry, "state estimation"?

Nick Rehm12:15 PM
VTOL solely for takeoff/landing is definitely desirable when runway space is limited

Nick Rehm12:15 PM
@Joe Stavitsky state estimation meaning measurement of a drone's orientation/position/motion in space relative to the world

Joe Stavitsky12:16 PM
@Nick Rehm interesting. It never struck me as a major problem but googling now it seems I was very wrong.

Sheridan-Tech12:16 PM
Do you think there is still room for small companies in the UAV space?

John Evans joined  the room.12:17 PM

Nick Rehm12:17 PM
I would say state estimation is 95% of the problem of deploying any sort of autonomy in a real-world environment. You may have seen those awesome autonomous drone aerobatics demos--looks closely. All of the state estimation is happening via motion capture cameras

pascal.fust12:17 PM
My point was, that in case the operation requires VTOL transitioning capabilities, aircraft design should take advantage of it, but rarely does

anfractuosity12:17 PM
Have you looked into controlling multiple drones, for things like lighting displays, that looks like it'd be very complex, but looks v. impressive

pascal.fust12:18 PM
....i.e., why should an aircraft be able to fly at low air speeds when it has VTOL units?

Nick Rehm12:18 PM
@pascal.fust ahhh, very true.

Joe Stavitsky12:18 PM
@Nick Rehm One thing that fascinated me was the choreographed big networks of tiny drones. If you've seen those, would you say the state estimation problem is more or less difficult for them?

Nick Rehm12:19 PM
@anfractuosity definitely looked into it, but gets expensive fast between a ground station, distributed telemetry equipment, and then all of the vehicles lol

kjansky112:19 PM
Why hasn't a partially buoyant either He or H2 assisted drone become a useful means to extend flight time with the reduced power requirements?

Nick Rehm12:20 PM
To my knowledge drone shows like that are using a combination of RTK-GPS which is a really good state estimate on its own, as well as ground-based sensors to further correct GPS drift

Nick Rehm12:20 PM
I could be wrong

pascal.fust12:21 PM
@kjansky1 most probably because you would need a good volume which results in extended drag

RichardCollins12:21 PM
@Joe Stavitsky Any measurements that help you know what is happening. Velocity, acceleration, rotation, rotation rate, vibrational FFTs. Voltages, currents, battery and fuel capacity, temperature, density, clear air turbulence. Lots of sensors available, and thousands more in development, in thousands of labs. But a mess, expensive and everyone for themselves.

@kjansky1 There are stationary airships proposed for months and years long "low orbit" satellites.

Nick Rehm12:21 PM
@kjansky1 wind :)

Joe Stavitsky12:22 PM
@Nick Rehm not sure if we're talking about the same thing, I meant the really tiny ones. Let me see if I can find video.

RichardCollins12:23 PM
Above 100,000 feet. But if you can make big ones, then tiny ones are possible too. You have to decide what things you like making, and who you want to work with.

pascal.fust12:23 PM
@Nick Rehm Which system of aircraft control (autopilot,etc.) would you see as most promising for future developments?

Joe Stavitsky12:23 PM
Not exactly it, but close enough

Joe Stavitsky12:23 PM

https://www.youtube.com/watch?v=hj_SBSpK5qg

YOUTUBE MAVLAB TU DELFT

Nick Rehm12:25 PM
@pascal.fust I'm not sure I follow, could you elaborate? Like what general components need more work, or what existing packages show promise?

kjansky112:26 PM
If only partially buoyant and streamlined drag would be insignificant if only low speeds but increased loitering time would be the requirement. Wind is not a factor because the drone is still powered. Goodyear is fully buoyant and it has no problem in maneuvering.

Nick Rehm12:26 PM
@Joe Stavitsky Something like that is probably doing very simple position estimation using an optical flow camera

pascal.fust12:27 PM
looking at open source solutions, would you think that ardupilot, pixhawk, etc. are most promising for future developments?

Joe Stavitsky12:28 PM
@Nick Rehm I've seen other projects where they are much more carefully choreographed, like in a square grid that undulates vertically

pascal.fust12:28 PM
Or do you see more elaborated systems at the horizon?

Nick Rehm12:28 PM
@Joe Stavitsky Look for the motion capture cameras!

Joe Stavitsky12:29 PM
@Nick Rehm so you mean outboard right?

Nick Rehm12:29 PM
ardupilot/px4 are here to stay as the 'standard' I think

pascal.fust12:29 PM
okay, thanks, a word of a professional

Thomas Shaddack12:30 PM
Random thought. 3d-scan the environment, use that as a fixed reference, and motion-cap camera on the drone itself.

Nick Rehm12:30 PM
But skynode is one of the first attempts to integrate more computing directly olongside the autopilot for more autonomy: https://auterion.com/drone-manufacturers/skynode/ So definitely something to watch out for in the near future

Robot Dogger joined  the room.12:31 PM

Nick Rehm12:32 PM
@Thomas Shaddack SLAM - simultaneous localization and mapping is basically doing this, using computer vision to build a live map of the environment and estimate the robots position within

Joe Stavitsky12:32 PM
I love this thing. The company also did a bird.

Joe Stavitsky12:33 PM

https://www.youtube.com/watch?v=nj1yhz5io20

YOUTUBE FESTO

Joe Stavitsky12:33 PM
They do industrial manufacturing robots, mostly

Nick Rehm12:33 PM
I've been dying to make a hovering ornithopter

Joe Stavitsky12:34 PM
@Nick Rehm I wanted to copy it, but they're cheap on specifics :(

Nick Rehm12:34 PM
how about a cyclocopter?

Joe Stavitsky12:35 PM
@Nick Rehm there's a whitepaper, if you want the link

Nick Rehm12:35 PM
oh yes that'd be awesome if you could share

Joe Stavitsky12:37 PM
@Nick Rehm my mistake, a brochure https://www.festo.com/us/en/e/about-festo/research-and-development/bionic-learning-network/highlights-from-2013-to-2014/bionicopter-id_33493/

Nick Rehm12:37 PM
on the topic of robot dragonflies, here's a little video snippet from some research I did in undergrad:

kjansky112:37 PM
About 275 million years ago dragonflies had a wingspan close to 30 in. or 2.5 ft (75 cm) with an estimated weight of over 1 pound (450 g), which is similar to the size and weight of a crow. Build that with robotics!

Dan Maloney12:37 PM
Had a natural ornithopter (hummingbird) buzz me the other day. Just hovered by my face for a few seconds before figuring out I wasn't food and flying off. An amazing design -- if we could only pack that much tech into such a small package.

Nick Rehm12:37 PM

https://www.youtube.com/watch?v=qQAatJg3tnc

YOUTUBE NICHOLAS REHM

Myself12:38 PM
Do you do any work on GPS-denied environments?

Myself12:38 PM
(Of course as soon as I hit enter, I realize anyone doing such work probably can't talk about it if so...)

Nick Rehm12:40 PM
@myself lol, yes I do and it is actually one of my main interests. For some reason, I just hate GPS even though it works for 95% of drone applications

Nick Rehm12:40 PM

https://www.youtube.com/watch?v=p8frNNYQNV4

YOUTUBE NICHOLAS REHM

Nick Rehm12:40 PM
GPS-denied motion planning ^^^

Myself12:41 PM
Oooo, that's open in a tab for later! Very exciting.

Joe Stavitsky12:41 PM
@Nick Rehm so once I get my 3d printer back I'd like to start work on propeller design, can you recommend any texts?

Nick Rehm12:42 PM
Gordan Leishman's heli aero book is basically the bible of rotor design

Myself12:43 PM
I keep feeling like powerlines are one of the hardest obstacles to avoid; trees tend to show up on lidar a lot more reliably. How is that dealt with in the real world?

pascal.fust12:44 PM
@Joe Stavitsky Do you think that 3D printed structures are strong enough for props?

Joe Stavitsky12:44 PM
@Nick Rehm and what about software?

Thomas Shaddack12:44 PM
resin printers, SLA, could be strong enough.

Nick Rehm12:44 PM
That is definitely a very real problem. Detecting powerline-like things is very hard. Maybe somewhere that AI/computer vision will fill in in the next few years?

pascal.fust12:45 PM
@Thomas Shaddack Depends on the resin material?

Discussions