⭐🏆⭐PRIZES: Racing AI Competition⭐🏆⭐ DEADLINE: May 15ᵗʰ 2024

When designing your AI car, you can use the playground as instructed previously, but you are not limited to it as long as you produce a compatible “.car” file. This extra freedom comes with a warning: changing car specifications like the max speed, acceleration, friction is not allowed and you will be disqualified if you attempt to do so. You are only allowed to modify the sensors configuration and the neural network. Feel free to ask anything on Discord, you can also submit early and I can say if something is allowed or not.
🚗SELF-DRIVING CAR COURSE (PHASE 3)🚗
• Self-driving Car :: Ph...
💬DISCORD💬
discord.gg/gJFcF5XVn9
🚗PLAYGROUND🚗
radufromfinland.com/projects/...
radufromfinland.com/projects/...
radufromfinland.com/projects/...
radufromfinland.com/projects/...
🏆PRIZES🏆
- Everyone who defeats me with their AI cars will get one month channel membership for free
- The top 5 will get a recommendation letter from me (on LinkedIn, for example)
- The top 3 can also make a video / blog post describing their solution and I will promote it on my channel
- The winner (best AI submission) can have a video call with me (zoom, teams, etc) where we discuss your submission; you can show code, diagrams, slides, etc. I will record the call, edit it and post it on my channel
🏆DRAFT PROGRAM🏆
If the winner will be a student, alumni, or faculty member of Karelia, Riveria or UEF, we can offer free access to the Draft Program® (if you have a business idea you will get 1000 € funding and coaching): www.draftprogram.com/en

[@backstabba]

This is important and would be nice to pin: *The race track won't have intersections, just curves, so the path is clear.* I have a few more questions (and I haven't followed either in P3): -can I only modify the characteristics of the sensors or also their number? -am I required to follow the rule of 'staying in the correct lane'? -will my car have to evade other competitors or general traffic or is it a track race?

[@Radu]

Yeah, the cars won't have to decide which way to go. There won't be intersections, or, if there are, I'll use the path finding / corridor techniques to give the cars the corridor to follow, not the entire map. So, no need to worry about that. The 'right_hand_rule.car' will be able to complete the race just slow (it's very careful and stays in its lane). - You can modify the sensor count, spread, length and offset. The playground linked in the description has options for doing that (you can also do it with code if you want) - No, in the race, lanes won't matter. But don't touch the road borders. The cars will get damaged then. I'm still thinking if I'll make cars respaun when damaged or if I handle the collisions so it just slows cars down when touching the side (to be less harsh). But touching the side of the road should be avoided nonetheless and training the cars not to do that is important. - The cars will be on the track at the same time, but there won't be any collision detections between the cars. They don't need to know other cars exist, basically, like in Trackmania: just focusing on the time trial aspect of racing. This video at 19:36 has some details about the race: kzbin.info/www/bejne/inysfHmomNuhrJYsi=U-km5s04aeFxyDzf And the video coming next week teaches how to convert the simulation into the racing game. It and upcoming lessons will clarify some things even more.

[@dastan331]

You are my best teacher! Thank you! What can i do for you?

[@Radu]

I'm just happy when somebody watches :-)

[@Android-17]

Was eagerly waiting for phase 3 but missed it after first lesson due to work. Will catch up now and take part in it🏎️ Thanks Radu for phase 3🚀

[@Radu]

Cool :-) good luck!

[@alihussnain8311]

❤

[@Radu]

❤

[@ChandrashekarCN]

💖💖💖💖

[@Radu]

❤️

[@disrael2101]

would love a video tutorial on how to use ml to train the car

[@Radu]

Will eventually happen in Phase 4. It's why I asked you to help me collect data by racing in the first lesson of Phase 3. But I will need a long rest after Phase 3 is done, so, Phase 4 will be later.

[@blackhawk1484]

Can I submit a .car for each map separately, or does my .car have to be good in every scenario, like maps 1 to 4?

[@Radu]

It's important for your car to work well in general: on as many scenarios as possible. On race day the track we'll be racing on will be new! (more details were given here at 19:36: kzbin.info/www/bejne/inysfHmomNuhrJY)

[@davidmurphy563]

Ohhh... I haven't been following the series at all (sorry!! I'm working silly hours 7 days a week..) BUT, I _think_ I can get this Sunday off - yaay! So I can maybe do this then. So, is it the phase 2 virtual world that I need? Would it be cheating if I found a way to write it in python? I can code js, but honestly, I just don't enjoy the language... JQuery is better but, still, meh. I would just code it from scratch but even with the same settings it wouldn't be a apples to apples comparison with the other entrants. I guess I could fire up a django / flask server or how about I just wasm it? Pyodide or whatever it's called should do the trick. Although getting it exported onto your local system might be more effort than it's worth... Could put it online but firing up a server on my day off doesn't sound like fun. Hmm... Might have to pass. I guess I could code it in js... I'll think about it.

[@Radu]

Actually, it's phase 3 (understanding AI) and you don't 'need' to code in any language :-) I made a playground where we study neural networks by manually adding nodes, links, weights and biases. In lessons 1 - 3 i teach how to make the logic for a nice car (that stays in it's lane, can turn, etc). That can be used as a starting point for the racecar. Lessons 4 and 5 are with the playground as well, but they teach how to respect other traffic rules or find its way to a given destination. The race track won't have intersections, just curves, so the path is clear. Of course if you want to use python or any other programming language to help you adjust the weights and biases, that's fine :-) You can take the file the playground generates, read it into python, modify the weights / biases in python somehow, and generate a new file with those. Good luck! And ask if you have any questions :-)

[@davidmurphy563]

@@Radu Sorry, I had to work Sunday in the end. :(

[@Radu]

@@davidmurphy563 :-) no problem!

[@maytan8782]

where is phase 3?

[@Radu]

Should be in the description.