After years of having the equipment laying around, I've finally begun to dive more into creating robots using lidar. So I'm curious, is there really machine learning involved (required) here? Or is it just saying - here are solid obstacles on either side, I need to stay N distance away from it while I travel forward.
@@blaircox1589 ⚠️ God has said in the Quran: 🔵 { O mankind, worship your Lord, who created you and those before you, that you may become righteous - ( 2:21 ) 🔴 [He] who made for you the earth a bed [spread out] and the sky a ceiling and sent down from the sky, rain and brought forth thereby fruits as provision for you. So do not attribute to Allah equals while you know [that there is nothing similar to Him]. ( 2:22 ) 🔵 And if you are in doubt about what We have sent down upon Our Servant [Muhammad], then produce a surah the like thereof and call upon your witnesses other than Allah, if you should be truthful. ( 2:23 ) 🔴 But if you do not - and you will never be able to - then fear the Fire, whose fuel is men and stones, prepared for the disbelievers.( 2:24 ) 🔵 And give good tidings to those who believe and do righteous deeds that they will have gardens [in Paradise] beneath which rivers flow. Whenever they are provided with a provision of fruit therefrom, they will say, "This is what we were provided with before." And it is given to them in likeness. And they will have therein purified spouses, and they will abide therein eternally. ( 2:25 ) ⚠️ Quran
This is really exceptional work. I love how your thought process is always generating the next possible improvement, and then you just keep pushing to refine your designs.
The most intriguing thing about this according to me is realisation that you could work leanly with data sets. Yes, within a certain dataset eg. maze (the square one for example) you want as many laps as possible, but for industrial purposes keeping the the amount of mazes down when you know what kind of mazes the robot will encounter should also avoid bloating the Arduino with unnecessary data. Also, truly amazing that you made a robot that could race faster by itself than you could race it manually, just by tweaking the motor speed. That goes to show what machine learning can do in terms of work optimisation, sort of like how the search function on a computer vastly outpaces any human manually searching for a document in an archive room. Thank you for making a video that clarifies so much with so little!
Great project. The reason it was able to handle the increased speed was twofold: 1) the control dynamics were similar enough at both speeds and 2) the sampling rate was high enough that the time delta didn't have an impact on the stateless prediction model.
🎯 Key Takeaways for quick navigation: 00:00 🤖 *Introduction to the robot and project setup.* - Introduction to a small robot with machine learning algorithms running on Arduino Uno. - Overview of the project's goal: autonomous navigation on a racetrack. - Mention of the steps to be covered in the video, including building the robot, creating a racetrack, data collection, processing, and a final race. 02:20 🤖 *Building the robot and the racetrack.* - Description of the robot's construction using an open robotic platform. - Explanation of using simple blocks for the robot's chassis and adding necessary components. - Improving traction on robot wheels with TPU tires. - Innovative use of cardboard for creating racetrack walls. 05:01 📊 *Data collection setup.* - Installation of a Bluetooth module and an SD card for data collection. - Explanation of recording lidar measurements and control labels while driving the robot. - Details about the data format and collection process. 07:10 🧠 *Processing and training the machine learning model.* - Discussion of feature selection to reduce data dimensionality. - Overview of experimenting with different machine learning classifiers. - Mention of using Python libraries for processing and training. 09:08 🏁 *Testing the robot's autonomous driving capabilities.* - Introduction to testing the robot's performance on various racetracks, including square and figure-eight. - Highlighting the ability to adapt to new racetracks with additional training. - Preparing for a final test on a complex racetrack. 11:40 🏎️ *Achieving high-speed autonomy.* - Surprising results as the robot successfully handles high-speed autonomous driving. - Discussion of motor speed settings and PWM signals. - Comparison of robot performance between manual control and machine learning algorithms. Made with HARPA AI
I wonder if you could set up virtual race tracks in something like Unity to collect training data. The benefits would be that you could control it with an actual controller and from first person ( so better precision ), and you could have much larger and more complicated tracks
Nvidia have created Isaac Sim for that purpose, it can even generate photorealistic images on the environment for training but just using a distance sensor like this should be easy to implement. It is made for training ML algorithms so can generate the training data and train the AI and can even do things like reinforcement learning, genetic algorithms or similar.
Very cool. I've been programming for a little over 40 years but I've never had the time / opportunity to delve into machine learning. Closest I've gotten is using AI services for photo processing. Now that it can drive itself it would be an interesting progression to give it memory of where it's been, building a map of the course it travels and being able to use that to plot improved trajectories for future loops. Just like we're slow when traversing unfamiliar territory but with repeated trips we can anticipate and optimize our course. You should be able to borrow from tech such as CNC path processing which can optimize acceleration / deceleration for curves and apply that to steering. Just an idea.
That is doable. But, a more efficient (and less fun) way of doing it would be to build a simulator. Use RL methods in a sandbox, train the model and load it into the robot and that's it!
Well done! This was a very satisfying video to watch. Well explained and I totally understand the thrill of building something that actually works in the end!
Very cool use of Random Forest to do edge avoidance and robot navigation. Another alternative for deploying such a model as C code is emlearn, which is a bit more actively maintained. Disclaimer: I am the maintainer
In general I just take inspiration on UA-cam to make my own projects, because it’s not exactly the way I would’ve done it. But for this robot, I would’ve done it exactly the same way if I had the idea, so I’m going to do it anytime soon ! Thank you very much for this cool video, and congratulations, that’s impressive !
Great video and content! Love how you bring us through the journey of your experiments and the tidbits of discoveries that is available oit there. BTW, little editing features like the ghosting effect really elevates your game. I must agree with other commenter, some of your voice recording suffers in quality (when in testing area, hard walls). It does not impact my opinion but, you are competing for attention against others. I hope you continue to push this project further. Maybe like an iRobot that travels throughout the house for guard duties or identify any new objects...
Wow! Thank you for this amazing video. I was always wondering how lidar data is being used to train machine learning models. Most of the ML tutorials are object detection using images.
Nice Work! It would be interesting to see if one could "simulate" the movements for a rectangular track, instead of training on the actual path. I would guess, it would lead to comparable results. If yes, then the advantage with simulation is that one can design more complex paths without actually building them - making the training process very efficient and robust.
Would it not be easier / faster to generate the features via simulation rather than you driving by hand? That way you can train on way more tracks (as they can be pseudo-randomly generated) and also have the simulation provide feedback on collisions when driven by the AI model.
You could used ordinary tracking a line on the floor or collition detection to teach the LIDAR to track. And you could even use that input while learning. Then you can remove that tracking device. It is not uncommon to use extra input while doing the learning.
If you have access to, or can get an Nvidia GPU then you could train it in a virtual environment and use the virtual environment to gather training data, then you only need to implement the ML model on the arduino. Edit: you don’t strictly need an Nvidia GPU and for simple ML like this could probably get away with just using your CPU but Nvidia has a lot of support for ML and robotics. Your best bet for really good results is to train it using reinforcement learning in a virtual environment. That way you can change the track much faster too.
Very cool project. Perhaps I would have approach the training of the model via software in a simulated environment as it would be way faster to collect data always in the optimal path.
Very cool. Any thoughts on how youd approach training a model for pan tilt camera object tracking that could out perform standard control algorithms like PID or MPC? I know I could use the data from the PID controller to train a model to do the same thing, but i wany the model to be better and feels like the only solution there is some type of deep reinforcement learning using real world setup
Can it be made to teach itself? Technically, it has a rangefinder on the top. In the square, for example, it shouldn't be running into the wall, when faced perpendicular to it. Is there a way to have it go through the course at low speed, be aware of when it completes a lap, optimize its path, then gradually increase its speed, as it becomes more confident? That could save a few minutes/hours of manual training, while also making full speed training unnecessary. You could have a routine that varies the amount of left to right distance bias, and after a few runs it could have complete familiarity with the track. Then you might be able to send a command to either hug the inside, outside, or prefer the center, while the robot is moving. Just a thought.
Could you have made a synthetic dataset for this using something like ray tracing on a 2d simulated race track instead of manually training it? I don't know much about lidar so maybe not.
I cam make some suggestions as a 5 year in a row roborace winner. Me Personally, i would put a switch on the robot that allows for automatic learning or performing, u can use the distance from the sensor as a valid/invalid logic for keeping the data or trashing it due to a crash (too short of a distance, mean crash), also try to add a preference on the robot as far of lefty or righty robot, doing so in can also navigate maze, and i usually add also a preference for the longer the distance it measure in a direction the bettere it is, but with all of this, to proper race u need to now how to access really low level on arduino or a more powerful mc. Anyway if it's your first time, not bad. Ps. With lidar won't be common but implement a system to detect when the signal is wrong (like wall to far or too close, or a reflective wall, or sometimes also really pointy corner can cause problems, like if u have an Y too narrow. Usually i do like, of i detec for x time too similar data, do a sweep around and check if Im stuck
A really interestin and amazing project. You are the Master. Nikodem, I need to do a ball robot collect. So, how to do for to adapt your experience on a ball collect robot arm by arduino? My idea is: - to put sensors on each axis (angle sensors) to move. Maybe 4 axis (minimum). - To move the arm manually. Or is better to use by BT like you? or in each movement to put a switch and when the movemt is fisnish to push the switch? - To write each sensor data on a SD. How to organize the datas on SD for to TinyML? - Use the TinyML to create the code "*.h" and then use it on the arduino. Some other suggestion for to adapt you project on my robot? And another option could be will be to use a "PixyCam2", it generate a lot of data could be interesting (size, position, reco👌gnise object, ...)
That would be a different type of machine learning. Here none of the actual learning is done on the robot, he just gathers training data and then trains the algorithm on a computer. What you suggest would involve training on the robot itself or setting up a separate line follower system just to gather the training data. Also line following isn’t that useful and can be difficult to do well. You can do the simple way of just using colour or reflection sensors but beyond that you have to go into machine vision which the arduino can’t handle. The best way to do it would be to set up a simulation and then use something like reinforcement learning to train the robot.
@@conorstewart2214 i never said that it should be done on the arduino. instead of driving it yourself, the robot does itself by following lines and save it on the sdcard as usual.
@@TechnopolisDotTV still that isn't the best option. You are just teaching it how to follow a virtual line essentially. It also involves making a separate line following system too that is just used for gathering data and for it to be useful for gathering data it would have to be a well tuned system for smooth and consistent movements, not just the jerky and slow versions that most people make. How you lay out the line would heavily influence what path the robot takes round the track. Reinforcement learning in a virtual environment would be much better and would let the ML decide on the path itself. You also want some inconsistency in training data and driving it around the track manually provides that, so it doesn't take the same path every time, with a line the path is pretty fixed.
(sorry, I am really interested on what you have used to concert python language to for Arduino and relative description but it is not in your comments, could you please give me a direction, thanks) found it in the link at your comment, thanks
Siemanko. Na wstępie musze powiedziec ze bardzo rzadko pisze komentarze, ale.. Twój filmik a bardziej projekt mnie powalił! =CZAPKI Z GŁÓW= pomyślałem zobacze co to za kolo, bo tak ładnie mówi po angielsku a tu okazuje sie ze jesteś z Polski.. Świetna robota naprawde! Czemu ja nie mam takich właśnie kolegów ;) pozdrawiam Marcin
Heyyy, This is pretty cool. Wouldn't adding an RL Algorithm(Maybe Q-Table or a Deep-Q) to this solve the problem of manually having to drive this around?
hello , if I want to measure the density of traffic on a road, what kind of system can I install? i want something low cost and functional. it should be a system that can transmit this density to another device and show the density. i would be very grateful if you help me. lidar sensor is used? which one do you think is advantageous?
Please can you tell me if the Arduino supports the amount of data from the lidar? because the arduino will have to control the motors according to the lidar data,.Really. I would appreciate your response.
Cool!, you mentioned that this is the first time you have used knowledge from the university. What are you currently studying? What kind of specialization?
New subscriber here! This is a fantastic project. I just came to this video from the one where you raced two of them, very cool stuff. I am working on a similar project using an RP1 LiDAR with a differential drive robot as a learning platform for things such as SLAM. I am not too familiar with ML. Do you have any favorite resources or project ideas to get begin learning?
Hey i am just wondering to work with the same project but i am thinking of a maze solving bot. maybe it is a great idea to work on one using arduino with a lidar sensor but i havent got the materials to do it yet. i think it would be good if you work on a video like that
How about running this as MicroPython or AdaFruit's CircuitPython directly on a more powerful microcontroller called the RP2040? It's the debut mcu from RaspberryPi. I'm going to power my keyboard PCB with that.
The ESP32S3 is better for ML, it has DSP instructions and has a higher clock speed and works with micro and circuitpython too. Also micropython is not a good language for microcontrollers, they are just too resource and performance limited. You don’t want to lower the performance even more by running an interpreted language on them. C is much better for microcontrollers, the programs require less storage space and they run much faster. Versions of Python used on microcontrollers should only be used for very basic prototypes or projects.
Of coarse. I would opt for software training though while normalizing the data to what the lidar module outputs. then you can do multiple training sessions in parallel then upload the finished modal to the robot for real world testing. If you want to do reinforcement training, you would need in excess of 30,000 - 50,000 revolutions around the track.
Since I’m not playing with such things my question is: Why do you have to train it? Since it got a radar just tell it not to come too close to something. 🤔🤷♂️ So easy 😜
Great project 👍 ❤ Can we use any simple sensor system for this ? To make it cheap and simple to implement. And to all the commentors who are suggesting to train on viral environment and then implement it on robot, no doubt we get the perfect path for driving but how this will react to actual sensor data ?
Even 6 could do as I am using 2S lipo battery which is 8.4V max. But it's better to use 18650 or a lipo instead of AA batteries. You can get higher currents and recharge them
We are Fatsa Science High School artificial intelligence and robotics club students. We want to improve ourselves in this field. We liked your project very much, we bought all the materials needed for the project and printed 3D prints. Could you please send us the circuit diagram of the project?
![Lp. Сердце Вселенной #60 РОЖДЕНИЕ ЛОЛОЛОШКИ [Финал] • Майнкрафт](http://i.ytimg.com/vi/YoR0pAV9FVQ/mqdefault.jpg)
More about the project: indystry.cc/ml-robot/
Happy making!
After years of having the equipment laying around, I've finally begun to dive more into creating robots using lidar. So I'm curious, is there really machine learning involved (required) here? Or is it just saying - here are solid obstacles on either side, I need to stay N distance away from it while I travel forward.
@@blaircox1589 ⚠️ God has said in the Quran:
🔵 { O mankind, worship your Lord, who created you and those before you, that you may become righteous - ( 2:21 )
🔴 [He] who made for you the earth a bed [spread out] and the sky a ceiling and sent down from the sky, rain and brought forth thereby fruits as provision for you. So do not attribute to Allah equals while you know [that there is nothing similar to Him]. ( 2:22 )
🔵 And if you are in doubt about what We have sent down upon Our Servant [Muhammad], then produce a surah the like thereof and call upon your witnesses other than Allah, if you should be truthful. ( 2:23 )
🔴 But if you do not - and you will never be able to - then fear the Fire, whose fuel is men and stones, prepared for the disbelievers.( 2:24 )
🔵 And give good tidings to those who believe and do righteous deeds that they will have gardens [in Paradise] beneath which rivers flow. Whenever they are provided with a provision of fruit therefrom, they will say, "This is what we were provided with before." And it is given to them in likeness. And they will have therein purified spouses, and they will abide therein eternally. ( 2:25 )
⚠️ Quran
can you create Tubercle + Toroidal version fan...?
This is really exceptional work. I love how your thought process is always generating the next possible improvement, and then you just keep pushing to refine your designs.
Haven't been doing robotics for a while and this is one of the coolest videos. I got recommended.
I’m trying to start bro it’s so cool. I just got my learners kit
@@sorryboss8550 where from
@@bofa722 anywhere, got mine from a store near me👍🏽
The most intriguing thing about this according to me is realisation that you could work leanly with data sets. Yes, within a certain dataset eg. maze (the square one for example) you want as many laps as possible, but for industrial purposes keeping the the amount of mazes down when you know what kind of mazes the robot will encounter should also avoid bloating the Arduino with unnecessary data.
Also, truly amazing that you made a robot that could race faster by itself than you could race it manually, just by tweaking the motor speed. That goes to show what machine learning can do in terms of work optimisation, sort of like how the search function on a computer vastly outpaces any human manually searching for a document in an archive room.
Thank you for making a video that clarifies so much with so little!
Great project. The reason it was able to handle the increased speed was twofold: 1) the control dynamics were similar enough at both speeds and 2) the sampling rate was high enough that the time delta didn't have an impact on the stateless prediction model.
On an Arduino! I am grabbing a LIDAR module as soon as possible. Thank you for the videos.
🎯 Key Takeaways for quick navigation:
00:00 🤖 *Introduction to the robot and project setup.*
- Introduction to a small robot with machine learning algorithms running on Arduino Uno.
- Overview of the project's goal: autonomous navigation on a racetrack.
- Mention of the steps to be covered in the video, including building the robot, creating a racetrack, data collection, processing, and a final race.
02:20 🤖 *Building the robot and the racetrack.*
- Description of the robot's construction using an open robotic platform.
- Explanation of using simple blocks for the robot's chassis and adding necessary components.
- Improving traction on robot wheels with TPU tires.
- Innovative use of cardboard for creating racetrack walls.
05:01 📊 *Data collection setup.*
- Installation of a Bluetooth module and an SD card for data collection.
- Explanation of recording lidar measurements and control labels while driving the robot.
- Details about the data format and collection process.
07:10 🧠 *Processing and training the machine learning model.*
- Discussion of feature selection to reduce data dimensionality.
- Overview of experimenting with different machine learning classifiers.
- Mention of using Python libraries for processing and training.
09:08 🏁 *Testing the robot's autonomous driving capabilities.*
- Introduction to testing the robot's performance on various racetracks, including square and figure-eight.
- Highlighting the ability to adapt to new racetracks with additional training.
- Preparing for a final test on a complex racetrack.
11:40 🏎️ *Achieving high-speed autonomy.*
- Surprising results as the robot successfully handles high-speed autonomous driving.
- Discussion of motor speed settings and PWM signals.
- Comparison of robot performance between manual control and machine learning algorithms.
Made with HARPA AI
“I am an algorithm I need more learning and training” ❤ cheers to you!😊
For all Polish people! I started a new polish YT channel: www.youtube.com/@prosteczesci
I wonder if you could set up virtual race tracks in something like Unity to collect training data. The benefits would be that you could control it with an actual controller and from first person ( so better precision ), and you could have much larger and more complicated tracks
Nvidia have created Isaac Sim for that purpose, it can even generate photorealistic images on the environment for training but just using a distance sensor like this should be easy to implement. It is made for training ML algorithms so can generate the training data and train the AI and can even do things like reinforcement learning, genetic algorithms or similar.
@@conorstewart2214That's really great 👍. Is it open to use for free?
At that point it would be possible to run ppo with a neural network instead
Watching this, I was shocked to realise that you don’t have a million subscribers. People are missing out!
Very cool. I've been programming for a little over 40 years but I've never had the time / opportunity to delve into machine learning. Closest I've gotten is using AI services for photo processing.
Now that it can drive itself it would be an interesting progression to give it memory of where it's been, building a map of the course it travels and being able to use that to plot improved trajectories for future loops.
Just like we're slow when traversing unfamiliar territory but with repeated trips we can anticipate and optimize our course. You should be able to borrow from tech such as CNC path processing which can optimize acceleration / deceleration for curves and apply that to steering. Just an idea.
This was awesome, congratulations!
I've used o-rings for tires for 3D printed wheels.
Good idea!
Really like the ghost version when comparing the speed. That was nice. Make a other one and see if you can train to pass slow robots.
Great job! It would be great to see this robot learning by itself by reinforcement learning
That is doable. But, a more efficient (and less fun) way of doing it would be to build a simulator. Use RL methods in a sandbox, train the model and load it into the robot and that's it!
Well done! This was a very satisfying video to watch. Well explained and I totally understand the thrill of building something that actually works in the end!
Very cool use of Random Forest to do edge avoidance and robot navigation. Another alternative for deploying such a model as C code is emlearn, which is a bit more actively maintained. Disclaimer: I am the maintainer
In general I just take inspiration on UA-cam to make my own projects, because it’s not exactly the way I would’ve done it. But for this robot, I would’ve done it exactly the same way if I had the idea, so I’m going to do it anytime soon ! Thank you very much for this cool video, and congratulations, that’s impressive !
Huge thanks for a really nice comment!
First time I see your channel, this is great! Awesome work !
Very nice robot and video! ❤
Great video and content! Love how you bring us through the journey of your experiments and the tidbits of discoveries that is available oit there. BTW, little editing features like the ghosting effect really elevates your game. I must agree with other commenter, some of your voice recording suffers in quality (when in testing area, hard walls). It does not impact my opinion but, you are competing for attention against others. I hope you continue to push this project further. Maybe like an iRobot that travels throughout the house for guard duties or identify any new objects...
Hi. You can make a smart vacuum cleaner with automatic cartographer on sd card?
Wow! Thank you for this amazing video. I was always wondering how lidar data is being used to train machine learning models. Most of the ML tutorials are object detection using images.
Had a look at the RPLidar DataCollector code, what is the purpose of setting the Lidar resolution to 240? Why not set it at 360 degrees?
Nice Work! It would be interesting to see if one could "simulate" the movements for a rectangular track, instead of training on the actual path. I would guess, it would lead to comparable results. If yes, then the advantage with simulation is that one can design more complex paths without actually building them - making the training process very efficient and robust.
Would it not be easier / faster to generate the features via simulation rather than you driving by hand? That way you can train on way more tracks (as they can be pseudo-randomly generated) and also have the simulation provide feedback on collisions when driven by the AI model.
Fascinating! Thanks for posting this video!
The coolest project ever on machine learning. So realistic
5:09 "I'm not an expert... I'm just learning and experimenting by implementing some machine learning in my projects."
Great channel, glad I found it.
You could used ordinary tracking a line on the floor or collition detection to teach the LIDAR to track.
And you could even use that input while learning. Then you can remove that tracking device. It is not uncommon to use extra input while doing the learning.
If you have access to, or can get an Nvidia GPU then you could train it in a virtual environment and use the virtual environment to gather training data, then you only need to implement the ML model on the arduino.
Edit: you don’t strictly need an Nvidia GPU and for simple ML like this could probably get away with just using your CPU but Nvidia has a lot of support for ML and robotics.
Your best bet for really good results is to train it using reinforcement learning in a virtual environment. That way you can change the track much faster too.
Can you explain little bit more how this can be done?
@@aayush212 your best option is to look at "OpenAI Gym" or Nvidia's "Isaac Sim" and "Isaac Gym".
Un crack el Niko! Gracias por compartir tus conocimientos!!! 👏👏👏
gracias!
What are these kind of breadboards called? the smaller ones he is using in the robot. like shown in 00:49 ?
Mini bread boards??
Very cool project. Perhaps I would have approach the training of the model via software in a simulated environment as it would be way faster to collect data always in the optimal path.
감사합니다. 열심히 공부해 보겠습니다.
Very cool. Any thoughts on how youd approach training a model for pan tilt camera object tracking that could out perform standard control algorithms like PID or MPC?
I know I could use the data from the PID controller to train a model to do the same thing, but i wany the model to be better and feels like the only solution there is some type of deep reinforcement learning using real world setup
Can it be made to teach itself? Technically, it has a rangefinder on the top. In the square, for example, it shouldn't be running into the wall, when faced perpendicular to it. Is there a way to have it go through the course at low speed, be aware of when it completes a lap, optimize its path, then gradually increase its speed, as it becomes more confident? That could save a few minutes/hours of manual training, while also making full speed training unnecessary. You could have a routine that varies the amount of left to right distance bias, and after a few runs it could have complete familiarity with the track. Then you might be able to send a command to either hug the inside, outside, or prefer the center, while the robot is moving. Just a thought.
Well done. Highly appreciated from Pakistan.
Keep it continue.
Could you have made a synthetic dataset for this using something like ray tracing on a 2d simulated race track instead of manually training it? I don't know much about lidar so maybe not.
I cam make some suggestions as a 5 year in a row roborace winner.
Me Personally, i would put a switch on the robot that allows for automatic learning or performing, u can use the distance from the sensor as a valid/invalid logic for keeping the data or trashing it due to a crash (too short of a distance, mean crash), also try to add a preference on the robot as far of lefty or righty robot, doing so in can also navigate maze, and i usually add also a preference for the longer the distance it measure in a direction the bettere it is, but with all of this, to proper race u need to now how to access really low level on arduino or a more powerful mc.
Anyway if it's your first time, not bad.
Ps. With lidar won't be common but implement a system to detect when the signal is wrong (like wall to far or too close, or a reflective wall, or sometimes also really pointy corner can cause problems, like if u have an Y too narrow. Usually i do like, of i detec for x time too similar data, do a sweep around and check if Im stuck
Teach me bro pls 😢
A really interestin and amazing project. You are the Master.
Nikodem, I need to do a ball robot collect.
So, how to do for to adapt your experience on a ball collect robot arm by arduino? My idea is:
- to put sensors on each axis (angle sensors) to move. Maybe 4 axis (minimum).
- To move the arm manually. Or is better to use by BT like you? or in each movement to put a switch and when the movemt is fisnish to push the switch?
- To write each sensor data on a SD. How to organize the datas on SD for to TinyML?
- Use the TinyML to create the code "*.h" and then use it on the arduino.
Some other suggestion for to adapt you project on my robot?
And another option could be will be to use a "PixyCam2", it generate a lot of data could be interesting (size, position, reco👌gnise object, ...)
Great work. You do already amazing things. And you are young. I wonder what you will do in a year or 5 or 10. Keep going!
Can you make the movement using a servo motor like the fire extinguisher robot in the contest on the Trinity University ?
Super cool. Great job!
Love it!
Pro Tip (i guess):
In Training Mode: Let the robot follow a black line instead of driving it yourself.
That would be a different type of machine learning. Here none of the actual learning is done on the robot, he just gathers training data and then trains the algorithm on a computer. What you suggest would involve training on the robot itself or setting up a separate line follower system just to gather the training data. Also line following isn’t that useful and can be difficult to do well. You can do the simple way of just using colour or reflection sensors but beyond that you have to go into machine vision which the arduino can’t handle.
The best way to do it would be to set up a simulation and then use something like reinforcement learning to train the robot.
@@conorstewart2214 i never said that it should be done on the arduino. instead of driving it yourself, the robot does itself by following lines and save it on the sdcard as usual.
@@TechnopolisDotTV still that isn't the best option. You are just teaching it how to follow a virtual line essentially. It also involves making a separate line following system too that is just used for gathering data and for it to be useful for gathering data it would have to be a well tuned system for smooth and consistent movements, not just the jerky and slow versions that most people make. How you lay out the line would heavily influence what path the robot takes round the track.
Reinforcement learning in a virtual environment would be much better and would let the ML decide on the path itself.
You also want some inconsistency in training data and driving it around the track manually provides that, so it doesn't take the same path every time, with a line the path is pretty fixed.
(sorry, I am really interested on what you have used to concert python language to for Arduino and relative description but it is not in your comments, could you please give me a direction, thanks) found it in the link at your comment, thanks
This is a very cool project! Machine learning on an Arduino! Imagine Teensy or ESP
Siemanko. Na wstępie musze powiedziec ze bardzo rzadko pisze komentarze, ale.. Twój filmik a bardziej projekt mnie powalił! =CZAPKI Z GŁÓW= pomyślałem zobacze co to za kolo, bo tak ładnie mówi po angielsku a tu okazuje sie ze jesteś z Polski.. Świetna robota naprawde! Czemu ja nie mam takich właśnie kolegów ;) pozdrawiam Marcin
Heyyy, This is pretty cool.
Wouldn't adding an RL Algorithm(Maybe Q-Table or a Deep-Q) to this solve the problem of manually having to drive this around?
hello , if I want to measure the density of traffic on a road, what kind of system can I install? i want something low cost and functional. it should be a system that can transmit this density to another device and show the density. i would be very grateful if you help me. lidar sensor is used? which one do you think is advantageous?
Please can you tell me if the Arduino supports the amount of data from the lidar? because the arduino will have to control the motors according to the lidar data,.Really. I would appreciate your response.
I recommend you making it so it will learn while riding by itself.
zajebist kanał stary, oglądałem aż w końcu mnie uderzyło "chwila chwila...ten akcent brzmi bardzo znajomo...XDD"
This awesome. I want to learn machine learning more than every
Amazing. I like what you're up to. Keep it up!
Super impressive! :o
How much please. OK to advise what parts to buy or used parts and where please..eg lidar. Have u tried used vacuum cleaner lidar too?
Cool!,
you mentioned that this is the first time you have used knowledge from the university. What are you currently studying? What kind of specialization?
ooo where did you buy your cutting mat?
Wonderful work!
Great work, super cool!
New subscriber here! This is a fantastic project. I just came to this video from the one where you raced two of them, very cool stuff. I am working on a similar project using an RP1 LiDAR with a differential drive robot as a learning platform for things such as SLAM. I am not too familiar with ML. Do you have any favorite resources or project ideas to get begin learning?
Dude this is so cool!!!
the final track should be a combination of circular, sharp edges & cross junction...
Is it possible to make the same project with multiple ultrasonic sensors instead of ladar
Loved that you needed algorithm, hehe
Hey i am just wondering to work with the same project but i am thinking of a maze solving bot. maybe it is a great idea to work on one using arduino with a lidar sensor but i havent got the materials to do it yet. i think it would be good if you work on a video like that
is it possible to replace the track with a room like a warehouse that has several rooms???
I really really REALLY want to know how you are controlling the arduino (I am assuming) with your phone interface
hello, can we do it with ultrasonic sensor ?? and if please make video
You have a sensor for penalty in case of the error?
Interesting and very inspiring! Thanks
Yery impressive!
A fun project. 👍
Good video for robotic knowledge
Thank you for helping me ❤❤❤
You data collection code includes Serial1, I thought Serial-3 were only available on the mega 2560, How are you making these work on an uno?
That's Uno R4 with the new more powerful microcontroller and you have more serial ports available
@@nikodembartnik Cool, thanks, I have just ordered one to play with.
How about running this as MicroPython or AdaFruit's CircuitPython directly on a more powerful microcontroller called the RP2040? It's the debut mcu from RaspberryPi. I'm going to power my keyboard PCB with that.
The ESP32S3 is better for ML, it has DSP instructions and has a higher clock speed and works with micro and circuitpython too.
Also micropython is not a good language for microcontrollers, they are just too resource and performance limited. You don’t want to lower the performance even more by running an interpreted language on them. C is much better for microcontrollers, the programs require less storage space and they run much faster. Versions of Python used on microcontrollers should only be used for very basic prototypes or projects.
hi sir i wanted to what language did you use for ML. is it c??
amazing, very good!!! i like you project.
can any one provide sources for setting up the robot (like tutorial thing) or any assistance
How can we use reinforcement learning instead of labelled data set
is It possible>>>?
Of coarse. I would opt for software training though while normalizing the data to what the lidar module outputs. then you can do multiple training sessions in parallel then upload the finished modal to the robot for real world testing.
If you want to do reinforcement training, you would need in excess of 30,000 - 50,000 revolutions around the track.
Love that how it performed
Since I’m not playing with such things my question is: Why do you have to train it? Since it got a radar just tell it not to come too close to something. 🤔🤷♂️
So easy 😜
Oh no, almost same as my graduation thesis😢
But I am using GP algorithm so is it ok..?
You are amazing, keep up
wonderfull thank you for sharing and good luck
Well done!
Hello sir, can you share how to convert Python to Arduino library, and how to apply it?
How long did it take you to collect the data?
can you create Tubercle + Toroidal version fan...?
Nice Work
How can I learn to do what you do?
Wait, how is possible that it can handle the circular path but not the squared path? Geometrically talking it should be much more easier 🤔
10:35 A switch statement would be more appropriate
Great project 👍 ❤
Can we use any simple sensor system for this ? To make it cheap and simple to implement.
And to all the commentors who are suggesting to train on viral environment and then implement it on robot, no doubt we get the perfect path for driving but how this will react to actual sensor data ?
Can i power the motor driver with 8 AA batteries?
Even 6 could do as I am using 2S lipo battery which is 8.4V max. But it's better to use 18650 or a lipo instead of AA batteries. You can get higher currents and recharge them
@@nikodembartnik Thank you for the answer. Does it matter wich TPU u use for the wheels?
We are Fatsa Science High School artificial intelligence and robotics club students.
We want to improve ourselves in this field. We liked your project very much, we bought all the materials needed for the project and printed 3D prints. Could you please send us the circuit diagram of the project?
Great Job Brother keep going
Great topic, thanks 👍