How Did I Use Python And Ai To Create A Fully Autonomous Drone? ArduCopter pixhawk and OpenCV.
HTML-код
- Опубликовано: 25 июн 2021
- This is how I created a fully autonomous drone capable of advanced autonomous missions using Python, Ai and Computer vision. This drone is capable of following a moving target fully autonomous and will be able to avoid obstacles! And even better I am going to explain to you how it all works and how you can create your own fully autonomous ai powered drone using the free open source Python code in the description below!
multirotorresearch.com/
Full Code: github.com/sieuwe1/Autonomous...
Bas: / bas-beckers-00a308163
Niek: / niek-van-dam-514711131
WHAT DO YOU WANT ME TO MAKE NEXT? LET ME KNOW IN THE COMMENTS! 
Наука
Thanks for sharing. Can you also post the hardwares/materials to build this drone including the sensors and everything? Thanks.
So basically this is what you need:
-- A drone. Basically any drone that has a pixhawk or similar flight controller that runs ardupilot wil work. But if you are building from scratch you could build something like this:
- Get 10 inch drone propellors and motors (t-motor is one of the best brands). A motor between 400 and 1000kv should work fine.
- 30a Esc’s. Any brand is fine but t-motor or hobby King make good ones.
- 4s lipo battery. 3s is possible but 4s will increase flight time a lot and also give the drone more power for faster following. Tattu makes good batteries.
- flight controller that can run ardupilot and is released after the pixhawk. The chineese clones of the pixhawk can be really bad sometimes. A more expensive unit but which I really like a lot is the cube orange. Otherwise a radio link pixhawk is also good. I would recommend buying a flight controller in a kit. This will make sure that you get everything you need like wires, speaker, safety button, telemetry system etc.
- gps + compass module. Get a radio link module if you are on a budget or a Sirius f9p from dronetek. Don’t buy a cheap unit. Gps and compass is one of the most important sensors for a drone. Getting a cheap one will result in crashes!
- for the frame a dji f450 or tarot clone of this frame is good. Going with quad motors is the easiest and cheapest.
-- beside a basic drone we also need some specific parts. These are:
- jetson nano 4gb module. (Dont buy the 2gb model it is not good)
- a 1080p csi camera. I like modules with the imx219 sensor a lot. Easiest would be to just buy that one.
- a step down converter like the LM2596 I used.
- a tf Luna LiDAR module from benewake.
- a 3d printer to print some custom brackets located on GitHub.
- adding a usb wifi module to the jetson will enable remote debugging of the system using ssh. Definitely recommend this.
That would be it! Good luck trying to make it. Definitely send me a video if you have it flying.
@@sieuweelferink6852Hey, where I live the pixhawk and similallars are out of stock. Do you think I would be able to crete something like that with a beta flight FC? I know they are more focused for racing but I wonder if I could create an autonomos drone like that. It is a FC Omnibus Betaflight F4 PRO V2 Flight Controller STM32 F405
@@sieuweelferink6852 Hey buddy Even I am working on the similar project using ML in it. Can you suggest me the best controller in the Pixhawk exactly so that I can get the documentation and really use it. And also I want to add the sensors to it like air pressure sensors etc. Is there a provision to add those sensor and if it is possible then which sensors are available in the market for it. your Guidance will be Valuable
He is a type of guy I am searching my whole life to be friends with.
suprisingly, robots hunting people down scares most people
This is the kind of youtube channels WE want to be recommended to 🙏.
I have subscribed
Thank for sharing this and the breaking it all down! I have been wanting to make my own from scratch as well and make it fully autonomous!
Oh man, this is genus. I watched another video on a similar project using the Jetson and it wasn't even close to as smooth as this solution. Very elegant, and it's python!
This is brilliant work !! Well done!
Nicely done! Don't give up sharing your content. Even if there are similar projects out there.
Very cool, this is great stuff man. Keep it up and thanks for sharing.
Looking forward to this!!
This is too cool, thank you for making it open source
Really good work man!
YOU ARE OVERPOWERED!!!
opened my eyes to a pretty cool hobby my dude thanks :D
Great job buddy, learned a lot🙏🏼🙏🏼🙏🏼
Brilliant video, inspirational!
great work. thanks for sharing. do upload more of your work. you are such an inspiration for drone developer.
Keep it up bro, well done.
Great project! Thanks for sharing with us
Thank you🙏
You're really a good kind and helpful young man👍
Cheers🕊️
Awesome project!
Amazing work
thanks for sharing, this is so wonderful
Very nice E46 man I really like it it's a coupe, I have the sedan (4 doors) 330i, I'm an embedded system engineer I do some PCB design for IoT devices, I'm also a ROS enthusiast actually I was watching some videos on PX4 MavROS and I found your video.
Keep up the good work bro you are amazing I really people like you
Really nice . Thank you.
Just amazing
Oh yes, that what i need😍
Amazing 👏👏👏
wow you re crazy, keep going!!!!
Amazing. You are genius.
Sieuwe you are great person ❤✌👋
Great job! Could u please refer to the sensor models? which rgb camera and lidar did u use?
so cool. thanks
Thank you!
Thanks for sharing
love it
Thank you very much for your efforts.
Thank you for watching!
That was fantastic man . Please upload more❤️
I will try as much as possible! But my projects are complicated and thus will take time. I will now focus 100% on the autonomous vehicle V2.0. So stay tuned for that!
@@sieuweelferink6852 great man. good luck.
What happened with this project? Did you stop working on it?@@sieuweelferink6852
Good Job Sieuwe
Thanks for the support!
nice job man...
Thank you for using 3s. My favorite battery size. :)
How were you able to calibrate both rotations and forward movement for the right PID control?
Hello Sieuwe, it is a great work. I have the M690L hexacopter from Align. It is a bit older one, so i intend to use an additional hardware/software to update the AI possibility of this drone. Is this O.K. that you could check what is possible with genuine flight Controller of this copter?(APS M). Is there a way to contact you (PN)?
amazing
Now way!
You should make a home security drone. That would be awesome!
Thanks for the video 🤙
That’s a really good idea! He should do that.
Is it even possible to simultaneously implement the mav condition yaw (for yaw) and mav frame body ned (for forward)?
Can we control pitch yaw and roll without using pid control in program?
nice project, when is your part 2 coming, (autonomus obstacle avoidance drone) ?
Hi. Where did you connect the Lidar to? To Jetson? Thanks
Hi Sir,
This is very great project!!
May i know the spec of motor you’re using in this project?. It is very much appreciated
I wish no one will come and buy you out becos we need you in these forums :)) !!
this is cool...:)
what if instead of jetson nano I transfer this information ground station (laptop) and perform algorithms and command back to my drone. Is it possible??
How hard would it be to switch from tracking mode to a "kill mode"? Where drone will collide with tracked target at full speed.
I like great🎉🎉
which library is used for drone control? it doesnt seem that you have used drone kit
please keep uploading
Uber Cool.
Can i do this project without a lidar sensor and by using a raspberrypi5
Engels zeer on point😂
I would like to see two or more drones follow the same target in synch with each other.
Are there any cheaper alternative for jetson nano board
WOW HAMZING!!!
Great job but you need more recognition 👏👏👏👏👏
Thank you for watching!
Can you please confirm that this will work on a PX4 stack as well .
i had Deja Vu @6:58😱it felt like i already watched that part of the video🤯
Can you please make a project on gps and navigation UGVs for agriculture!🙏
Great work and solution! Carry on, and you could have a great future, by doing this as your future job!
On question, if possible. Do you think it could be possible to integrate a library of objects the drone has to track and follow? For example Cars,, Bikes, People or equal, or is therefore AI needed? I mean something same like the Autel Evo Max is offering?
Yes definitely. The model I used is trained on the imagenet dataset. So basically the drone could track every object which is in this dataset. I think there are over 50 different objects. I just simply filter for only the object type person. But changing that line to a list will enable what you want.
Is it possible to complete the task with more cheap flight controller?
Thanks for the video. I am really interested and i want your guidance in making a full drone if you don't bringing your help forth, I would I appreciate
genius
hi my drone need to detect a box and pick it up its similar to ur project can u help me out
I was literally looking for this exact thing for delivering packages to my friends in my neighborhood
You are clever🎉🎉🎉🎉
Where to get the same drone?
You should post a full tutorial
Great job This is something which is very useful in industries.I have been planning to build like one of these for autonomous tracking.
Could you please suggest me the PC configuration required to run a yolo model at 30FPS for custom object.
Thanks in advance
Jetson nano 4gb is fine for tiny yolo model. Otherwise look at jetson Xavier. Can run at a couple 100 fps!
@@sieuweelferink6852 how do you make the image run on 25fps i tried everything (im a beginer) and the best i got is i open image and than close it and repeat and its slow,garbage,stuppid,annoying.any sugestions?
are you still using the PS4 Stereo camera?
I need your help regarding autonomous flying drone without gps pls help n reply
Can this code run on a raspberry pi?
hi! are you still working on this? part 2?
Noice
Bro can you make a complete AI powered fpv drone building from scratch it would be alot of helpful for me please bro I want that so badly
Nice work man need to try this my self. Can you make tutorial fron no drone to fully automated drone. Building and programming
Hi so it is really easy. First build a drone with a pixhawk flight controller. Just search online on how to do it. Then mount a jetson nano and rgb and tf luna lidar on the drone. Install all software on the jetson nano and run main Python program.
@@sieuweelferink6852 thank you so much you are very helpful
Why did you connect lidar to jetson nano but not to PX 4? In this case, it would be possible to connect the Jetson to the Pixhawk via a serial port. And you don't need a USB to Serial converter. Data from the lidar can be obtained using the Mavlink protocol
Did he connect the lidar to Jetson?
1.What courses, what structures one should use, to make a drone fully autonomous?
2.How to start this work? How did you started? I am completely a beginer in drone programming, even i have no idea what micro-controller is best to use, and how to control the peripherals attached to the micro-controller, can you suggest any channel, any source or any helping websites, from where you prepared yourself for this outstanding work?
3.Please let me know where to learn about the drones and AI for drones, Thanks in Advance.
Hi I think most of your questions will be answered in the video! It will be ready saturday. Just as a sneak peak you need a drone with pixhawk flight computer. Also get a embedded ai processor like a jetson nano for the ai stuff. This is how I made my drone in the video!
sir which country u from,very good
Thanks for sharing. Could you post a new line follower drone using pixhawk?
a line on wall with arrow direction to move up down left right n yaw to go behind the wall
How long did it take u to master python?
hes got the jetson nano
Great man! What about HOME position update in INAV, or any other soft. to actual transmitter position, when transmitter is moving?
can you explain the code?
Hi @sieuweelferink6852, I love what you've done here. I'm aiming for a similar project which includes a net launcher for a final year engineering project. I've run into a dead end trying to develop an autonomous drone under a betaflight system. The plan now is to flash ardupilot to my existing SpeedyBee F405 V4 and pretty much run your code. However, I want to offload the processing to a ground station PC and not use a Jetson. Could you provide any tips to make this work seamlessly? The idea is to have identical functionality to your system, but tracking drones not people, and having a pc process the tracking and following code.
I think you need rtmp to get the video feed from drone and apply the object detection to detect person and then use the tracking code
I'm a subscriber just today and i will watch your video and share it and i can send you what I'm need, i just bought TELLO DRONE and i will like to program it using free open source opencv.. Like you are doing
Tello drone is a really good choice for simple opencv based projects! You can probably change our code to use with Trello drone and have good results👍 only you will have to change lidar distance to use for example bbox area as distance value
@@sieuweelferink6852 You are great.. Please I'm at other continent here where things for technology are very scanty but we are managing small resources, here even university have not resources except we check through Internet and i will be glad to be learning seriously from your channel...
Apart from RUclips do you have any plat form again?
Really amazing job sir you have done..
Hi I only upload on RUclips and sometimes explain my projects on instructables .
@@sieuweelferink6852 Thanks
Dear sir, i will try to use your code and if there is problem along the way, please i will be glad to get your help sir🙏
hI! im interested in using your crowd analytics project as a foundation in a project me and my classmates are currently trying to develop. is there any way i can contact you if we have questions? emails or discord calls?
Hi you can just ask your questions here so everyone can use the information i give you!
Great job! Will it work on raspberry pi 4?
No. The fps will be too low. Only 4 frames per second is possible. I tried.
Please upload a Drone update 🙏
why did you decide to use Jetson Nano instead of raspberry pi?
A rpi can not run a CNN like mobilenet in real time. Rpi can do something like 1 to 5 fps. A jetson nano can doe 30+ fps. More fps enables smoother control of the drone.
How can the drone collect data, i mean pictures of people and store it in PC?
You can just use the code already there. Jetson nano has a sd card slot where you can store picture on. Just change the code after the human filtering part
@@sieuweelferink6852 Thanks, i will try and after i will get in touch with you
why do you need the lidar assuming you can get the distance based of the a bounding box size?
Bro are you ok ?