I RACED My Self-Driving RC Car
Вставка
- Опубліковано 30 лип 2023
- I built my own self-driving RC Car and raced it against other RC cars. It's definitely FASTER than anything you've seen in my previous video. Get ready for crazy overtaking and defending.
If you haven't seen the first two videos featuring this self-driving RC Car, you can check them out here:
- • I Made a Self Driving ... (part 1)
- • Building the FASTEST S... (part 2)
Raceline Optimization Code: github.com/CL2-UWaterloo/Race...
Code ran on the RC Car (Will clean this code sometime this weekend...): github.com/CL2-UWaterloo/f1te...
My notes:
- stevengong.co/notes/F1TENTH
- stevengong.co/notes/Raceline-...
Credits
Thank you to Venkat’s group at Clemson University for providing the F1TENTH vehicle: sites.google.com/view/armlab-...
My professor’s lab: yashpant.github.io/group/
----------
MY SOCIALS
🌐 Website: stevengong.co
✍️ Blog: blog.stevengong.co
📸 Instagram: / stevengongg
📱 Twitter: / stevengongg
☕ Buy me a coffee: ko-fi.com/stevengongg
----------
Subscribers at the time of uploading: 20.7K
----------
ABOUT ME
I'm Steven, a university student currently studying Software Engineering at the University of Waterloo. I like to build cool things with code. I started this UA-cam channel to share my love for learning and engineering with the world. I do everything ranging from electronics, CAD, control theory, programming, 3D printing, physics, and design.
My ultimate goal is to be able to make anything that I want. I'm working towards this by building increasingly difficult engineering projects that push the limits of my knowledge. Consider subscribing if you would like to follow my journey and learn alongside me as a student of the universe!
#stevengong #selfdrivingcar #rcracing - Розваги
Hey everyone, I'm excited to finally share this video with you all! I actually finished filming this video back in May, but the editing ended up taking sooo long (like ~150 hours spread over 2 months). I had so much footage to go through, but I hope the end result made it all worth it. If you want to build your own self-driving RC, you can check out some of the links I put in the video description. Until next time 👀
Nice video, you did great! I have an idea concerning the localization issue. My idea might sound crazy at first, but I think it will work on a race track. What if you don't have a map? Just try to go as straight as possible. You would need a Gyro so you can detect unwanted spins. If the wall is far away, go fast; if it is near, slow down. That won't result in the fastest route, but I think it will work. The biggest downside is that it won't work in a normal environment like an office. I have programmed and competed at robotic competitions myself, so I really like seeing other people's ways to approach a problem. Keep going!
I just noticed something about 5:56 minutes in. My past life I was a driving instructor and the number one thing new students do, is not look far enough ahead. They typically end up looking not far off the front of the vehicle, within 2-3 car lengths. That's really close and dangerous because it doesn't provide enough time to process what is occurring and to react. So we reinforce how to look far ahead - very far and to use pereferal vision to be aware of everything in between. Your car is driving like a new student - it is only aware within that white box, so it's looking directly in front of the vehicle.
It's possibly a sensor limitation, but is there a way to feed your car long range information? Keep what you have as that is likely higher resiltiuon data, but can you augment long-range data as well so it can begin processing what is approaching, sooner?
I make a self racing drone if you are interested we can do something, it's an original design. I'm using a jetson i'v been doing the project for years.
you should make drift for sharper corners, more rotation and style points
Don't use RRT, use ERRT[1,2] which is intended for exactly these problems; I published it 21 years ago and it has been used by RoboCup (robot soccer) teams ever since. Section 4.3 in [2] has a good description, including plan-head simplification (pretty much required for any sampling planner) which is a generalization of what you were doing at 6:55, see [2] page 82 top paragraph. Hit me up if you have questions.
FWIW, [2] Chaper 5 solves the problem you don't know you have yet, and has also been used in RoboCup for years by at least 3 teams.
[1] Bruce, James, and Manuela Veloso. "Real-time randomized path planning for robot navigation." IEEE/RSJ international conference on intelligent robots and systems. Vol. 3. IEEE, 2002.
[2] Bruce, James Robert. "Real-time motion planning and safe navigation in dynamic multi-robot environments." Ph. D. dissertation (2006).
A high gloss basketball court was a cruel choice of venue for this competition. Great job, Steven!
en.wikipedia.org/wiki/Spherical_cow
I feel they could also use a MUCH better track boundary, that doesn't get pushed around, and isn't something thats very difficult for anything digital to see
Ye, the competition track felt kinda unprofessional / no shits given, especially compared to the time ppl put into their cars. They could have used gym benches for barries or somthing if they wanted a low effort solution.
@@ledorfYeah, the track was tight and made of *round RIDGED tube*… 🤦
lack of traction = emotional damage
I gotta say, I'm a 31 year old software engineer who also does robotics as a hobby, and you are absolutely killing it. I hate to sound like an old man, but being this far into the field when you're so young is amazing. Keep kicking ass, your videos make me excited to get back to work on my own machines!
I am 39 and hadn't even started the hobby o am dreaming about...so i am the old man here :))))
I'm 50 and have no idea! Do I win? 😅
Haha looks like you're the winner for now, but you're young enough that someone could still beat you! Glad to know we're all here together!
I am 93 and just started programming. It’s a great way to spend my time.
Yah dude, fucking TRUE! 37 years robotics and engineering nerd reporting in, you're doing great! Keep it up. 😄
Doing this solo is f-ing nuts. To me this is far from failure. From one engineer to another, you did really well!
Dont be discouraged Steven, you did your best effort and you were all alone in it. The track was pretty janky but one may argue that the algs should be smart enough to update the map on the fly. I think what could have made a difference is to practice on a map similar to the competition map since it was much narrower and might require some tuning to get working right, as well as corner cases like reversing once you have realized you are stuck. The tire grip also played a huge part and I think sticky tires would have helped. WD-40 degrades tire rubber to make it pretty sticky. But regardless, great effort and great content, thank you for sharing.
I'm wondering if the car could localise off of the ceiling (what with it being a static feature), but use the realtime LIDAR input from track walls/other cars to deal with collision avoidance - like how a human following a GPS is technically route finding using triangulation with satellites, but is avoiding crashing (mostly) by looking out of the windows.
You probably don't even have to update the map on the fly but just increase the measurement (and motion) model uncertainties in your positioning algorithm so that it trusts the map less and is able to model those inconsistencies. Out-of-the-box lidar-based stuff is most likely tuned to trust the map way too much.
To be honest, that track looks so simple that it could be driven around with a local planner only (no global map). Of course you would lose your race lines and probably lose track with some catastrophic odometric failure, such as spinning, but with low enough speed the local planner should be able complete laps.
In any case, super interesting video. Failures and difficulties make it interesting. Keep up the good work.
If this was done professionally then yes you would expect the cars to be able to deal with track variations but this presumably isn’t the first time they have done this competition, so they know what to expect in terms of car capabilities and they should make the track in such a way to maximise competitiveness. It’s not much of a competition when only one or two teams are able to reliably go around the track. In pretty much any other kind of racing, the track will not change, so at this stage, this competition shouldn’t be any different.
Changing to a rigid track would make it more competitive and it would change the winning criteria from whose algorithm can deal with changes in a janky track, to whose algorithm can go around the track the fastest. The best algorithms would still have the speed advantage.
@@conorstewart2214thats basically the tiny mouse robot challenge where the track is just tape on the floor
easy-to-move track boders seems like a massive oversight for something like this
Or a good way to test algorithm robustness? Most interesting robotic applications happen in more or less dynamic environments.
I’m sure there is a level of fairness, for example - masking tape outlining the course so the marshals can put it back to its original shape.
the map changing should not be a random variable, if its going to change at all it should be a new track for each run to test how a car learns a track
i race rc cars at our local race tracks and i often help build/change carpet tracks and dirt. we race with transponders, all data loged and i must say it cant be anywhere close to fair with the track changing as everyone goes around.
@IlariVallivaara except these are RC cars, worth hundreds, not NASA rovers worth billions.
you aren't going to catch a marlin with a trout rod
Man, i'm surprised at how uncontrolled the environment was when I first saw it. Its insane what you achieved with your resource and time constraints as a solo.
4:54 The easiest way to adjust the gears is to put a slice of regular paper in between them and then tighten them.
yupp
The last race human versus code was insane. I hope you keep going and compete again soon
The barriers / track limits changing lap to lap seems a fairly steep challenge on its own. Track limits are permanent in real life... imagine if every time a race car left the track the officials had to re-paint the white track limits line... during the race. Heckuva job facing some very difficult challenges!
ETA: My approach as an official would have been to at least put painter's tape on either side of the barrier to give a solid reference point when resetting. Another approach would be to anchor the tubing every "X" inches. I'd suggest painter"s tape under the barriers (to protect the surface), then glue the barriers to the tape. FWIW, painter's tape has a pretty good shear adhesion strength, and is often used in combination with CA glue to anchor pieces when woodworking
That wouldn't have helped the poor and tight design of it in the first place. Those RC cars have a lot of mass and power, they would rip tape
Dude, the "Max v paper strategies" bit made me spill my drink, nice 💯
I thoroughly enjoyed watching and getting to be part of this journey! And although you did not win you still gained a ton of experience, knowledge and "rivals". Maybe a shorter wheelbase would have helped with the car's turn radius, and if their barriers were semi-permanent the localization could potentially have worked more reliable, but hindsight is always 20/20. Again watching your journey has been fun and today's video has not been disappointing xD
Awesome presentation of the material! Very interesting, informative, humorous! A great illuminated idea! Good luck and thank you so much for sharing!
Super nice video, great to see your progress, well done !
Absolutely loved watching your progress from the first design to this one!
You did excellent! You were right on pace with the other drivers, and this was your first shot.
On top of that your video editing is top notch!
Been watching the build process and am excited you got to run! Good job on the car and the videos! Thanks for sharing your journey.
Good job, my guy. I had a lot of fun watching you progress. Thanks for the video
Man, that was a brilliant set of videos! It was so interesting watching you solve problems. I was so thrilled to watch each step of your journey and can't wait for your next mission.
Loved the ending too - so funny!!
Absolutely amazing. What dedication to set out and achieve. My hat is off to you. Keep it up. Just by being in it you are a winner.
Dude this was amazing! Especially that ending race, clipped audio and visual mashups 🙌🏽
wow...... this is amazing, well done for the good hard work you have put into the race preparation , the experience you got from this race will go a long way in helping you prepare for other races
I have never been so emotionally invested in a project before! This was not a fare race! A real fare race has a fixed track that does not change, with the main focus being speed and overtakes. You rock man, I'm am so proud of you!!
super fun to watch, interesting, inspiring
editing was definitely worth it
This was an incredible effort. Well done! So enjoyable to see such progress and thinking.
I love that you're making robotics fun and accessable. Keep up the great work!
This whole project was super inspiring! Great achievement participating in this competition. Wish you the best!
I am an old man into technical and mechanical stuff who got into RC airplanes. Veteran flyers seeing my frustrations at the field (crashing) gave me good advice and this “Everybody crashes”. After a few months of crashing and epiphanies I finely got it right in my mine and plane. You are well on your way to many epiphanies in your life, just remember, everybody will crash. Enjoy the ride.
as software developer and rc hobbyist, all I could say is I enjoyed your unit testing which consisted of yourself running behind the car lol amazing job btw
As a robotics engineer and a person who has competed in such competitions I can relatet to all your feelings and I thoroughly enjoyed this series! You are so good man, you not only put all your effort into this project but also actually spent the time making these videos to inspire others which is even better than winning the competition. Good luck with uni and your future as a robotics engineer my guy!
Great video, the edit was hilarious yet amazing. The project is astonishing. Just keep on going.
amazing project and editing, the ending was so cool to see too, way to go!!
You did a great job recording this whole steps. I thought I was in the competition with you. I really enjoyed following you along for this whole thing. Thanks.
Great job man! This is seriously one of the coolest engineering project series I’ve seen. It sucks to put all those hours in and be a bit disappointed in its performance at competition, especially when it feels so out of your hands, but either way your final design was beyond impressive! Not to mention the best video creation on the side that we get to enjoy
So sick!! Great work dude
You are so hard on yourself. What you have done is more than amazing. Congrads!
Great production quality on your video. Was very entertaining to watch. Nice work. :)
Good JOB mate, just love your work!!
I participated in FRC robotics for almost a decade. This is so similar haha, I love seeing you compete and try hard and seeing the difference between the teams with shit loads of money vs the teams with almost nothing, keep going! You will crush them next year!
such a good video, keep up the good work mate, ill be watching the whole series!!
so incredibly awesome! love this! would love to see a tutorial on all of this
that is amazing, dude! Keep going!
Loved following along your maker journey for your racecar. Winning isn't everything. You did a great job!
Holy shit! Really great work! Congrats, hats off!
Dude, everything about the video is fire. The editing is o on point. Subscribed for sure
Beyond incredible what you've achieved there. Mind-boggling in an environment like this!
Very good video, thank you for uploading! Waiting for the next update
Thank you for making these videos, I really enjoyed watching them.
Dude, you did great beginning to end. Keep your head up and keep going!
Pretty cool!
The las part of the video where you went against the self-driving car, *was epic!!*
this is so cool, I can't wait to see more!
Most excellent performance! Great video and great programming! Well done!!
I liked the race you put at the end really put in perspective how fast the cars were going.
Great combination of technical content and humor. Well done!
So cool and funny at the same time. Awesome work 👊
Amazing! Great ingenuity!
Incredible work, keep it up !
I love it the way you create story and suspense. Definitely I will see the code aswell.
Eventhough you might not have won, the video is really entertaining, you (probably) learnt a lot and inspired a lot of people (like myself). Great job!
great video! cool to be along for the ride (literally and figuratively)
Don't be disappointed, you did a great job. Awesome project and nice video editing skills. Keep it up 👏
brillant work and video editing.
Your enthusiasm and levity makes the subject that much more enjoyable.
Haha, great video!! I really enjoyed the robot race and your human vs robot race too!
Damn, I just wanted to check out the car but I stayed for the whole video. Great project and race, but also amazing directing and montage!
It was really fun to watch your journey
Super cool! I'm a Master student studied SLAM. During my time in school, I've worked on mostly solo projects, so I can relate the frustration of working alone. The cool thing is, I actually learned the most and was able to do all of them by myself!! Keep it up, love the content!
Thank you so much for doing this content. I'm from portugal and I started this year studying ros2 from the internet (we don't talk about this subject at the uni) and this is exacly what makes me study every day.
This video was fantastic! You have a great future in front of you!
Respect for trying. I hope you learned allot and had fun too. And one against teams of four. And also my respect for sharing you story. Your experience is a victory.
What a great video. Super well edited. Good story. It will be interesting to see how it goes next time.
It's hard going into such a big event with stiff competition when you don't have any experience. There are so many variables that it is difficult to succeed. You did an excellent job though. I was hoping that you and some of the other competitors would put together a more solid track and do your own race. I hope you continue to work on projects like this or similar and continue to upload it to this channel. The content has been great.
I waited so Long for this Vid
Thank you 🙏🏼
Steven, you're awesome! You did all of this work by yourself and you held your own against well funded large teams. Way to go man!
You have my respect for the will of competition and creating this awesome video!
You did an awesome job! I can't wait for more developments. The people that put on the race, do you know if they will improve the track moving forward or is that 'part of the charm'?
Nice job Steven :D
Dude, that's an amazing work!!! Keep on going like this!!!! Im sure you'll be the winner next time!!!
Brilliant video. Very entertaining.
Love to see my Penn Engineering boys winning :)
Fantastic job to all participants - you all did a fantastic time driving (heh) this field forward that day!!!!!
that was awesome and inspiring one... i dont know nothing about this even though it was interesting one.... keep up with your works and all the best . god bless you
This is absolutely sick, huh? Earned a sub here bro
to improve localization you could add 2 things: accelerometric vector persistence and match based rejection for your ladar telemetry. If you can confirm with ladar a position with certainty > 80% then make that your position in the map. from then on use the accelerometer to move your simulated position in the map whilst scanning with the ladar. any time you get a 80% solution with you ladar make that your new position.
This is basically how I made networked cars drive smooth back when I made unity games by having them constantly simulated clientside where I updated the velocities of the rigidbodies as well as static positions.
Godspeed Steven! smashed that sub harder than your racer smashed that trashcan
You did great Steven!! Great talent and for sure will win your next race! Love your videos, hope we meet again!
You did great ... learning and application of the learning are the greatest prize.
Amazing job. Sorry about the localization problems. The race with yourself was awesome, and seems to indicate that if they did same thing in a building the race results might be much different.
The super max portion 😂😂😂😂 I'm done hahahahah, I absolutely love it. Was interested in your content as soon as I saw the thumbnail but now I'm subscribed as well :)
All right smart guy and I am impressed. To see how it began and now look at it go.
don't be sad man, you've done a very good job and you have tried your best !
The racing and overtaking was incredible !!
Great video!
awesome, it’s all about the friendships you make!
I thoroughly enjoyed the final montage
Really interesting video. As someone that put an ESP8226 on a remote control car, and added a web app on iphone with tilt for steering, I'm in awe about the complexity of your project. Amazing. Keep going 👍
Great video! I don't know anything about autonomous racing, but I do Formula SAE, and this was incredibly interesting. Also great editing, kept it entertaining.
The final scene had me guffawing out loud. Great job!
Loved the jabs at Crashtappen. Great video.
Your editing has improved so much, great job dude
thanks zac :))
Great channel!! I'm finishing my PhD in Autonomous Vehicle Perception and I'm loving your content. It's really hard to find that kind of content on UA-cam, and you make it really good. Congratulations on your channel and the advances in your project. It's not easy to cover all the environments that the vehicle can be in, but its results are really impressive. Have you considered using a camera to help the "vision" of the RC car? Love your channel, hope you have greats achievements.
yea, i’m working on a few things for that :))
It's hard and disapointing enough to be eliminated, but you managed to make a video out of it and post it here. I wish you the best for your future projects !