Training the car to do self-parking using a genetic algorithm
TL;DR
In this article, we'll train the car to do self-parking using a g...
For further actions, you may consider blocking this person and/or reporting abuse
finally a post that is not about react :O!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Haha :D But! The Evolution Simulator is still built on top of React :D
This is probably the sickest post I've seen on this site.
Just brilliant! Out of curiosity, how long dit it take you to write this post (ideally, can you give elapsed time - to mature ideas - and actual time spent on writing) ?
Thanks a lot for taking the time to write this insightful article!
The issue is that the work on the article has happened in a "spare" from the main work time. So, it took ~5 month in a background to learn about Genetic Algorithm basics, learn about Three.js basics, implement the Evolution Simulator, experiment with parameters, optimise the algorithm and performance. And after that it took ~2 weeks to write the article.
Wow, this is a fascinating read.
Thanks for kind words!
Genetic algorithm is so interesting. I'm curious, on such example, do you spend more time on the genetic algorithm side or on the simulation side? I'm just wondering if there are some framework available for developing such solution, to have out-of-the-box things like you have below your simulator? Also, shouldn't we have the parking spot distance as a sensor? How would the car know to move forward or backward when not having a free spot in it's sensors?
Julien, I've spent most of the time to come up with the Evolution Simulator. The Genetic Algorithm basics andwriting the article took approximatelly ~1 month of background work.
The issue is that the work on the article has happened in a "spare" from the main work time. So, it took ~5 month in a background to learn about Genetic Algorithm basics, learn about Three.js basics, implement the Evolution Simulator, experiment with parameters, optimise the algorithm and performance. And after that it took ~2 weeks to write the article.
Current implementation of the parking function is really primitive, so it hardly adapts to the parking lot that is different from the one that you see in the Evolution Simulator. It also hardly adapts to the new initial car position. But, still, I belive what trained cars will do is they will move beckwards all the time until they find a free space. But again, to have a more sophisticated behavior of the cars we would probably need to use a Multilayer Perseptron as a brain function instead of over-simplified linear equation.
Thanks, those domain are so interessant! Are you working in this field?
Yeah, the domain is pretty interesting. No, I don't work in this field. It is just a hobby for me.
Awesome work.
Thanks!
Really cool, thanks!
Brilliant post!
Glad that you liked it!
Wow, this was fun, and I learned some things!
I think you meant "stand still"
Fixed the text. Thanks for checking!
Dude! That's an awesome post! It's the most complete that I have ever seen! Congratulations & thanks so much for it!
Thanks for such a response!
Very nice and informative. Keep them coming!
Gald that you liked it!
Hello do you mind giving the roadmap to learn about it?
The article itself supposed to be kind of a roadmap about applying a genetic algorithm to a bit artificial but yet pretty fun task :)
Great article @trekhleb , when I was younger!, I had done some practices to schedule students and teachers times with PHP using GA.
Agian, thanks for sharing.
Thanks for the feedback 👍🏻