Autonomous Raspberry Pi RC Car Mad Max Style

[@ebswift]

After using tape for the track I have these cars absolutely ripping around, this video was an early test. More to come.

[@nwordpass5960]

Autodesk inventor!! love it

[@rickslick4462]

Cool project, I wonder if you put tape down for the racing lines, and added something like a g force/accelerometer too record the direction and maximum G's pulled before the car spins off on a bend. That way you could make a nice little spread sheet of max speed around different types of bends without coming of the track.

[@ebswift]

You're correct in your thinking for sure. I'm not sure about g-forces and cornering, but a lot of work is being done with accelerometers, encoders, lidars, ultrasonic sensors etc., around the donkey car project and other similar projects. Since this is AI after-all, it is a pattern machine, and behaviour is simply a response to sensory input so pretty-much anything goes. Have it predict stock market prices while it's racing ;). If you join up to the Donkey Car slack channel you'll be able to see all the discussions around these things, there are also various bloggs with info too. I strongly recommend building a car if you have any kind of interest in this area, you can do it a lot more easily than what I did if you follow the Donkey Car formula exactly.

[@wayneswan3092]

I'm hoping to build one capable of 50+ mph, and handling jumps.

[@ebswift]

This was definitely fun to get working I hope to get up to higher speeds and build a better track. The chain is good but doesn't allow the car to make a mistake and get back onto line. Stay tuned for improvements and new projects.

[@viacheslavk3587]

Hi! Could you give some advise on how many laps did you have to finish by yourself for training purpose?

[@ebswift]

10 laps is minimum to get the cars driving pretty well. I keep adding to that training to get them driving better.

[@viacheslavk3587]

@@ebswift thank you a lot! One more question if you don't mind - which kind of network architecture have you choose for this?

[@ebswift]

@@viacheslavk3587 I have the cars connected to an eduroam network using a service account, I don't maintain the network architecture. The training PC is then able to collect the files from the cars using FTP. If training during a STEM activity you need a fast GPU otherwise the training downtime will be too lengthy. I use a GTX1070 which completes training in a few minutes. My oldish 12 core CPU takes over 30 minutes.

[@nguyenmanhcuong6968]

Hello, can you show me how to do it

[@ChristopherKendrick]

What 3D printer do you use?

[@ebswift]

That cage and camera holder parts were printed on an UpBox +. Now I'm running a Funmat HT and Lisa Pro. I've seen cheap 3D printers on youtube that seem to do a great job so you don't need anything too fancy for most things.

[@cold_fashioned]

what chassis / base car kit did you use?

[@ebswift]

These are 1/10 ECX AMP MT, still very happy with them, but had to replace out steering servos on both after burning out fairly quickly. photos.app.goo.gl/YyOImHZWe4VAHfiV2 Not the cheapest chassis to start off on for sure, but they are decent. Fish around on the donkeycar slack channel and you will find a lot of talk on different chassis options, in particular trying to match the magnet car internationally.

[@cold_fashioned]

Thank you for the quick reply! I just joined the donkeycar channel, so I'll search around there. I really don't have time for a new project, but this still looks interesting and fun.

[@mousetrapdiy7927]

good!