I made a self coding autonomous for v5
Interesting. Could we see code or could you explain how it codes itself. That statement is very vague on its own and I am sure there are many people, including myself, that are wondering what a self coding autonomous is and how it works.
I Too am wondering. Is it like artificial intelligence
It uses print console to print python code. I can’t tell you much.
Then why post at all?
If you are talking about coding an autonomous coding skills run by logging all the driver’s moves and then replaying it… already been done over the years, and explained on this forum.
I hope you can be more descriptive about your approach to enable other teams to learn from your experience.
Sorry for continuing to ask but I m very interested in this. Do you input coordinates and it outputs code to drive to those coordinates? Or is it more like AI like @Brainiac was asking about where it generates the best route by itself?
It prints gps cords in python Code
Not seeing a lot of evidence to support your claim here.
Ive seen it before where you drive a route many times and it gets the average of that route then turns it into auton.
This is actually a cool concept. I don’t know if I would do it during this season but would definitely try it out on a small sensor bot over the summer.
Do you plan on using this AI generated code for competitions?
Yes we do plan on using it
I’m gonna assume everything you claim is true, because it is both reasonable and quite cool. How do you plan to use it? My team has used AI every year up until this one, because this year’s game, at least in skills, appears to be very match load and brute-push oriented. The only viable option I could see for AI is path planning via registering the position of triballs from a vision sensor to effectively score all the matchloads in the goal, but the vex provided firmware locks down the vision sensor so heavily that I find that very challenging to accomplish.
I understand your perceived need for privacy and a competitive advantage, so no need to say anything specific; however, I think your concept is very intriguing but potentially over-engineered or even impractical.
Best of luck!
This sounds like Rewind which has been around a long time. The idea is to regularly record positions while driving the robot and then exporting those positions into a program that follows them. In the past it has been done by logging the encoders on the wheels, but using the GPS may be a new spin.
Teams have had mixed success using these tools. The pro being it’s easy but the con is that it’s difficult to make corrections on specific part of the path. The PID and pathing libraries seem to be the preferred navigation systems.
It would be interesting to link a pathing library with the GPS coordinates. Perhaps even add field elements as either obstacles or point values. Then let AI optimise the most efficient method for scoring.
Sounds good, but devil is in details. Specifically, in how you handle the errors due to variability of each run.
Here is an interesting prior discussion on this topic: