Learning High-Speed Flight in the Wild (Science Robotics, 2021)
Vložit
- čas přidán 5. 10. 2021
- Quadrotors are agile. Unlike most other machines, they can traverse extremely complex environments at high speeds. To date, only expert human pilots have been able to fully exploit their capabilities. Autonomous operation with onboard sensing and computation has been limited to low speeds. State-of-the-art methods generally separate the navigation problem into subtasks: sensing, mapping, and planning. While this approach has proven successful at low speeds, the separation it builds upon can be problematic for high-speed navigation in cluttered environments. Indeed, the subtasks are executed sequentially, leading to increased processing latency and compounding of errors through the pipeline. Here we propose an end-to-end approach that can autonomously fly quadrotors through complex natural and man-made environments at high speeds, with purely onboard sensing and computation. The key principle is to directly map noisy sensory observations to collision-free trajectories in a receding-horizon fashion. This direct mapping drastically reduces processing latency and increases robustness to noisy and incomplete perception. The sensorimotor mapping is performed by a convolutional network that is trained exclusively in simulation via privileged learning: imitating an expert with access to privileged information. By simulating realistic sensor noise, our approach achieves zero-shot transfer from simulation to challenging real-world environments that were never experienced during training: dense forests, snow-covered terrain, derailed trains and collapsed buildings. Our work demonstrates that end-to-end policies trained in simulation enable high-speed autonomous flight through challenging environments, outperforming traditional obstacle avoidance pipelines.
Reference:
A. Loquercio, E. Kaufmann, R. Ranftl, M. Müller, V. Koltun, D. Scaramuzza,
"Learning High-Speed Flight in the Wild",
Science Robotics, October 6, 2021
PDF: rpg.ifi.uzh.ch/docs/Loquercio2...
Code & Datasets: github.com/uzh-rpg/agile_auto...
For more info about our research on:
Agile Drone Flight: rpg.ifi.uzh.ch/aggressive_flig...
Drone Racing: rpg.ifi.uzh.ch/research_drone_...
Machine Learning: rpg.ifi.uzh.ch/research_learni...
Affiliations:
E. Kaufmann, A. Loquercio and D. Scaramuzza are with the Robotics and Perception Group, Dep. of Informatics, University of Zurich, and Dep. of Neuroinformatics, University of Zurich and ETH Zurich, Switzerland
rpg.ifi.uzh.ch/
R. Ranftl, M. Müller and V. Koltun are with Intel Labs
vladlen.info/
Voiceover: Pete Edmunds | British Voiceover british-voiceover.co.uk/
Music Credits: scottholmesmusic.com under Free Creative Commons License - Věda a technologie
Imagine using these to search the woods for a lost child or hiker. You could send out a hundred of these to search and map the forest with infrared as well.
Probably controversial considering the ethical implications concerning privacy in different manners, but also as a way to find a runaway convict or something similar. It's always a little scary with the capabilities of a government or any other powerful group that's able to use this tech in a way that's not ideal. Either way, this is a fantastic breakthrough and I can't wait to see it in actual use!
Would be possible yes. I fly these kind of drones as a hobby and I think under the so called "experts flying" I'm quite average. The only problem I would foresee with this is the range from home/ starting point. Woods and forest areas might not have good 4g connection so you might need to provide your own link to be reliable/ not dependent. Although this drone can fly itself I imagine you want to know the status of the drone at all times. (Where it is flying, if it is still flying, maybe even it's video feed of the drone)
With your own radio link this is very limited especially if you go inside the forest where trees are in between you and the drone.
For myself with quite good gear in can fly for 500m/1km into a dense forest. But as from 500m I definitely get link issues (bad image). 1km is maximum. This is also on flat ground no hills etc. which might makes things worse. I could image more professional gear to go further (maybe with big directional antennas) but that might extend it to 2km's through a forest.
Skynet scout
Actually, this was an intetion by ETH Zurich ... find lost hikers and fallen trees in the mountains. :)
Or you know, the government using it to find anyone they don't like.
You know that as soon as your example is possible, my example will probably happen too...
It's a solid and useful work! Amazing!
Very cool and amazing project!
Is the very first part of the video, result of the study or it is just a motion graphics for marketing?
I am asking, because the first part of the video looks to good to be true?! a pro is flying the drone which is recording the video and they are passing very tight gaps. But after that, in the rest of the video, recordings are from a stationary camera and the drone passes easier obstacles?!
What a groundbreaking effort
Comercial drone part was so funny 😂
The background voice went from normal student lvl to Discovery Channel lvl.
I have a couple of drones able to run this , off board or onboard.
Looking by the drone specs to reproduce it.
So impressive!
so the network is running on the drone's hardware?
Nice work!
Am I really the only one who thinks of the Star Wars chase scene in Endor and is excited that there's finally a real world solution for it?
no. i was thinking that too.
Who the hell would dislike this video!
Amazing!
a truly impressive result
Hi from Turkey
Hello
great project!
amazing !!
Fantastic!
Potential slaughterbot swarms just got really agile. Yaaaaaay.
Hunting Osama ? Or …. is he hunting us?
Awesome 😃😃😃
Amazing
Put another way, you're not going to outrun or escape. Unless you pretend to be a tree.
Between that, biological weapons, knowing our location at all times, with both normal and musk's satellites, cameras everywhere, infrared, heat visors, they know what you think and want, there's no way of escaping nowadays.
yeah, these are like missiles that can fly in a forest..
Nah.. you can use an infrared camera or a multispectral camera to determine if someone is under camouflage..
Tomorrow's drones will flawlessly navigate complex urban environments to deliver munitions to randomly chosen civilians
Not so random.
@@catharticcathexis4527 people like us :)
ok that might be true but whenever people talk about the military applications of drones....i just think that it has to be done in secret...you cant use a drone because you're scared you will get shot...because they will just shoot the drone
Harika bir çalışma.
One day they hunt for humans with these crazy inventions
Bilim teknik dergisinden gelenler
Really cool. I am just a bit nervous this will end up with military applications. I hope I am wrong.
you are almost certainly not wrong
Of course it will
Military will always be first
duh.. thats what there hoping for.... ever seen the price of military contracts?
Still waiting for the crash outtakes
I'm curious about how did you track drone and record the drone video. If someone did it. I think this guy is really the best expert. hahaha
I think it's the drone mentioned in the video. So there are two drones.
@@arcamon67 I did not get your point, is the yellow drone in the beginning different than the red which was used for the study?
@@imannabiyouni3006 yes I think so.
nice way to send a c4 freedom care package to the target enemy
Sadly thats probably already been done. Flying mines.
Next step Hunter killer raptor/Dragonfly drones that hunt down drones and snatch them out of the air to capture or disable....or destroy.
And the robotics arms race continues.
Hopefully enough of these will find use in intelligent hands.
PRIVELEEGED LEARNING AGHHH
Wudi!
This could be a good watchdog compagnon for female runner in park/forest
So where the hell is Brian laundrie hiding!??
I saw an ewok
So basically you can't run or hide lol
Now imagine these with explosives roaming the forests of Ukraine for hiding Russians