I also worked with the A1 RPLidar for my project, a stair climbing vacuum robot, and I also managed to get data from the LIDAR with my Arduino Mega, but the problem is, that it is not that much data like it is when you use it with ros2 and a raspberry pi or something like that. That is why I made my tolerances for measurement bigger to be more safe.
By using the actual bot instead of simulations, you’re capturing the true, dynamic responses in real-world scenarios, which will undoubtedly enhance your bot’s performance. Incorporating a penalty for each collision is a brilliant way to raise the stakes and teach it to navigate even better. I'm genuinely inspired by your work-your progress has motivated me to step up from software to hardware development. Keep pushing forward! With this level of dedication, I have no doubt your racing cars will be achieving remarkable results in no time. Amazing work!
You always provide super great video content/quality (the music was great and timed really well), and your projects are awesome! Keep up the great work!
Awesome! Just finished an advanced ML course for my masters in CS. I found it to be very interesting. Surprised seeing the capabilities in an Arduino. Got my wheels turning about a possible masters project. Great work!
Really cool to see this data-oriented way of improving the system and its performance. This kind of work is really critical for real-world embedded ML usecases, and this perspective is often missed by a lot of the educational material out there today. Cheers! Jon, creator of emlearn library, a tool to deploy Random Forest et.c. to microcontrollers
I just came across this. Would be interesting to see the performance compared to a neural circuit policy or liquid time constant algorithm, since they are supposed to perform better with smaller datasets and smaller model sizes. A python library exists that provides templates for tensorflow and pytorch, but unfortunately not sklearn.
It seems the slow robot tries to avoid colliding with the overtaking faster robot. Maybe you can train it to ignore if the distance to a (near) obstacle increases. But maybe that is not important.
Nice! Now what a cool robot would be is a Vector Robot with a brain using a Jetson Nano. Never got around to it, but wanted to build a larger scale version of the Vector Robot and have it explore the house to find small objects to pick up and take home.
I am using kbest to select the dimensions that are the most correlated with the label so the algorithm will get rid of all the readings in the back that do not contribute any important data to the final control label. This part is a little bit more explained in the previous video so check it out: ua-cam.com/video/PdSDhdciSpE/v-deo.html
Am I correct in assuming that the control loop basically works "stateless", in the sense that the racing machine is not really aware of it's own position, it just computes the steering signals based on the LIDAR input? If so, the obvious next step would be to create a proper environment map, then navigate in that. You can do that using machine-learning as well I think, but there are static approaches as well(see also ROS). Awesome project!
Рік тому
Will you do like tesla using camera instead of lidar?
No Sponsor? subscribed, liked and Loved! Really nice Video again. I Like It. (Just have enough of pcbway as a sponsor all time and raid shadow legends😂)
Why not also control the drive speed to a certain extend? Make the robot slower if it gets to close to an obstacle...
Рік тому
Hi i tried to make a digital version of you ML model, i managed to collect data, but still it does not start to drive, will you make a tutorial to train and start predicting data?
R4... Great use of it, congrat! Shame about poor connection, have you consider connecting to homemade PCB? Less noise I hope. (Use the R4 connected to PCB, shorter link)
Hey superb video. I also aspire to be a robotics guy but don't know how to start, can you please make a video or something so that I can start this awesome journey ❤
Make a exclusive video about the sensor, cable, pinout and library so you can help de community as you said there is almost none material about it. Use a very simple and representative title so it gonna show on searches and you get a lot of views and subs.
Why wouldn't you use a deep reinforcement learning algorithm (like PPO or DQN)? you first need to make a simulator, train your model, then use it in real life. this will be much more better
As mentioned in the video I'd rather have fun driving the robot in a race track than spend weeks writing the simulation software. Is it a better approach? Definitely. But do I want to do it? Not now, maybe in the future :)
LIDAR (light detection and ranging) in general is a method of measuring distance with light but in this case it refers to a device similar to a radar but it uses light to measure the distance between the objects.
I noticed it, too, but I consider it a trivial complaint. It was quite easy to listen to, as far as I'm concerned, and not worth his time. But if you want sound panels and other superfluous embellishments, I suppose you could contribute to his Patreon account.
I don't think it's that bad. Previous video was indeed not the best when it comes to audio but here I think it is not an issue. I will try to improve it with time, I am not a fan of the look of sound panels on the walls so hopefully just adding more furniture will help as well and maybe some curtains hang close to the camera at the time of the recording.
Incredible!
Hi!
Thanks
I also worked with the A1 RPLidar for my project, a stair climbing vacuum robot, and I also managed to get data from the LIDAR with my Arduino Mega, but the problem is, that it is not that much data like it is when you use it with ros2 and a raspberry pi or something like that. That is why I made my tolerances for measurement bigger to be more safe.
By using the actual bot instead of simulations, you’re capturing the true, dynamic responses in real-world scenarios, which will undoubtedly enhance your bot’s performance. Incorporating a penalty for each collision is a brilliant way to raise the stakes and teach it to navigate even better. I'm genuinely inspired by your work-your progress has motivated me to step up from software to hardware development. Keep pushing forward! With this level of dedication, I have no doubt your racing cars will be achieving remarkable results in no time. Amazing work!
You always provide super great video content/quality (the music was great and timed really well), and your projects are awesome! Keep up the great work!
Thank you!
Cześć! Uwielbiam twoje filmu, bardzo mnie motywujesz, dzięki ci za to!
To push this project further and further to this point, really cool and impressive!
Thanks a lot!
Awesome! Just finished an advanced ML course for my masters in CS. I found it to be very interesting. Surprised seeing the capabilities in an Arduino. Got my wheels turning about a possible masters project. Great work!
Great! Do the robots have constant speed? Or could one robot drive slow behind the other and then accelerate when there is a big enough gap?
Really cool to see this data-oriented way of improving the system and its performance. This kind of work is really critical for real-world embedded ML usecases, and this perspective is often missed by a lot of the educational material out there today. Cheers! Jon, creator of emlearn library, a tool to deploy Random Forest et.c. to microcontrollers
Wow, I’m so excited for the new video!
Indeed, I like your videos very much!
Very cool!
Just curious about the open pit/death trap/stairwell thing in that space...
Is it just unfinished construction?
I just came across this. Would be interesting to see the performance compared to a neural circuit policy or liquid time constant algorithm, since they are supposed to perform better with smaller datasets and smaller model sizes. A python library exists that provides templates for tensorflow and pytorch, but unfortunately not sklearn.
Check out the first video: ua-cam.com/video/PdSDhdciSpE/v-deo.html
Awesome project!
Wondering how many times they fall of the stairs.
It seems the slow robot tries to avoid colliding with the overtaking faster robot. Maybe you can train it to ignore if the distance to a (near) obstacle increases. But maybe that is not important.
You can make them as a micromouse to solve the maze the fastest
your a great creator i should have subbed sooner
:)
the noise is in those breadboard pins, my friend.... solder, solder, solder ;)
wow amazing, the autonomous overtaking is 👌👌
Thank you!
It surprises me the algorithm hasn’t picked this video up yet
It surprises me too :(
I really thought it will do better. At this point I can totally admit that I do not understand the algorithm at all!
Love this! Keep em coming
Which LiDAR module would you use if you did this again today? Are there better one today for DIY stuff?
add cat ears
/\_/\
( o.o )
> ^
Nice! Now what a cool robot would be is a Vector Robot with a brain using a Jetson Nano. Never got around to it, but wanted to build a larger scale version of the Vector Robot and have it explore the house to find small objects to pick up and take home.
This is truly awesome man
6:56 is a great solution but won't they capture false data?
I am using kbest to select the dimensions that are the most correlated with the label so the algorithm will get rid of all the readings in the back that do not contribute any important data to the final control label. This part is a little bit more explained in the previous video so check it out: ua-cam.com/video/PdSDhdciSpE/v-deo.html
i am do with some kind things with lider sensor and this is very informative
I love this channel.
Am I correct in assuming that the control loop basically works "stateless", in the sense that the racing machine is not really aware of it's own position, it just computes the steering signals based on the LIDAR input? If so, the obvious next step would be to create a proper environment map, then navigate in that. You can do that using machine-learning as well I think, but there are static approaches as well(see also ROS). Awesome project!
Will you do like tesla using camera instead of lidar?
Where we found a robotics kit
Hey, are you using arduino uno R4 minima or wifi? I don't have enough money for a Wifi but I can get a minima. What are the major differences?
One of the robots is using minima the other one wifi. The microcontroller is the same but on wifi you have additional esp32 for wifi and bluetooth.
oh, thank you. It was silly of me to ask before watching the video completely.@@nikodembartnik
No Sponsor? subscribed, liked and Loved! Really nice Video again. I Like It. (Just have enough of pcbway as a sponsor all time and raid shadow legends😂)
Why not also control the drive speed to a certain extend? Make the robot slower if it gets to close to an obstacle...
Hi i tried to make a digital version of you ML model, i managed to collect data, but still it does not start to drive, will you make a tutorial to train and start predicting data?
Just subd, This is an inspiration
R4... Great use of it, congrat! Shame about poor connection, have you consider connecting to homemade PCB? Less noise I hope. (Use the R4 connected to PCB, shorter link)
Hello Nikodem, any chance you could upload the modified code to talk to the rplidar?
I will try to update github repository next week and make a nice readme that this project deserves :)
@@nikodembartnik Terrific, thanks mate
link to buy the structure
Hey superb video. I also aspire to be a robotics guy but don't know how to start, can you please make a video or something so that I can start this awesome journey ❤
Great work ..! do you have an circuit diagram of the model can you share it.
Wouldn't it be better to use ESP instead of arduino and send the data live to the server for training?
UNO R4 WIFI also has wifi so you can do that without esp
dang thats sick! good job :D
Giga wideo. Moja dziewczyna ma tak samo na nazwisko :)
This is very good ❤
It's more like you have two children playing😍
Make a exclusive video about the sensor, cable, pinout and library so you can help de community as you said there is almost none material about it.
Use a very simple and representative title so it gonna show on searches and you get a lot of views and subs.
dlaczego nie masz barierki na schodach hahaha
badass!
Awesome.
Needs pyrotechnics
Why wouldn't you use a deep reinforcement learning algorithm (like PPO or DQN)? you first need to make a simulator, train your model, then use it in real life. this will be much more better
As mentioned in the video I'd rather have fun driving the robot in a race track than spend weeks writing the simulation software. Is it a better approach? Definitely. But do I want to do it? Not now, maybe in the future :)
@@nikodembartnik Ah I think I have that part of the video, but I really liked the robot
What's a LIDAR?
LIDAR (light detection and ranging) in general is a method of measuring distance with light but in this case it refers to a device similar to a radar but it uses light to measure the distance between the objects.
@@nikodembartnik thanks. Is this second one found in autonomous cars?
hey bro, great work. can you make a MICROMOUSE robot. it would be great if you can build one(consider it as a challenge😉)
Why you’re using original Arduinos instead of cracked ones, you can buy few Raspberry Pi’s on that money
To support original creators. There are no cheap clones for their new advanced boards.
Captions, pls
Please do something with your audio, you are recording in big room without any sound panels and its hard to listen, otherwise i like your videos.
I noticed it, too, but I consider it a trivial complaint. It was quite easy to listen to, as far as I'm concerned, and not worth his time. But if you want sound panels and other superfluous embellishments, I suppose you could contribute to his Patreon account.
I don't think it's that bad. Previous video was indeed not the best when it comes to audio but here I think it is not an issue. I will try to improve it with time, I am not a fan of the look of sound panels on the walls so hopefully just adding more furniture will help as well and maybe some curtains hang close to the camera at the time of the recording.