Donkeycar 4.4 released with tons of new features, including path learning (useful with GPS outdoors), better Web and Lidar support and lots of bug fixes.
github.com/autorope/donke
See it race at on Dec 10
meetup.com/diyrobocars/ev
Follow
DIY Robocars
@diyrobocars
DIY autonomous cars, from 1/16th scale to go-karts to full-size. Racing, hacking, and more. DIYRobocars.com
DIY Robocars’s Tweets
We are already biting our nails in anticipation of the #NXPCupEMEA challenge! 😉 Did you know there are great cash prizes to be won? Form your team now: nxpcup.nxp.com
3
11
UC Berkeley's DIY robocar program
2
3
The next autonomous car race at will be on Sat, Dec 10.
Thrills, spills and a Brazilian BBQ. Fun for all ages!
Spectators: meetup.com/diyrobocars/ev
Competitors:
2
6
22
Points for style. racing in at thanks for taking the video.
1
2
6
3
5
26
Presenting: the Hacky Racers Robotic Racing Series in collaboration with #PiWars. Find out more and register your interest by 1st November! Due to take place in Cambridge, April/May 2023 - mailchi.mp/04d13eb3b64e/p
9
16
There will be three classes at this event: A4, A2, and Hacky Racer! A4 and A2 are based around UK paper sizing and existing #piwars, , , and bots/racers should all be eligible. Hacky Racers scale will be for full size Hackys! 2/3
1
1
1
Show this thread
1
Calling all UK based folks interested in , , , and similar robot racing competitions! are looking to run a robotic racing series next year, we're getting plans together now but follow me for updates
5
15
🏎️
After hours of video editing, I'm happy to share a best of my Twitch videos on learning to race with RL.
🏎️
Each part is now condensed to 20 minutes, so you can learn to race in one hour only now =)
Code is open source ;)
youtube.com/watch?v=jXVUig
cc
9
50
Fully autonomous wipeout in the corkscrew at Laguna Seca today. Got some tuning to do; this car is very fast, but very easy to spin, and the controller seems to overcorrect. Hope to get some good laps tomorrow once I figure out exactly what happened here
8
6
96
Show this thread
Modelling oversteer and understeer
1
7
I use Foxglove Studio for my visualization needs, writing my datalog files using their MCAP library. I've also used Plotjuggler in the past, which now also speaks MCAP! It's a wonderful tool and I highly recommend.
2
8
29
Here's my car's view of that race. About 8.4 second lap times for laps 2 and 3... both of us utterly destroyed the ~9.9sec track record. That was awesome.
1
8
32
Great racing against today at ! Pretty cool to both break sun-9s at the track today I think I got very lucky with the cones today when it mattered. Thanks as always for amazing venue. (My car is black)
11
40
216
Show this thread
1
ran a huge number of hyperparameter tuning experiments yesterday; now I can train a new policy, far with better quality, in 15 minutes instead of several hours. Future iteration speed is well worth the time investment; now I can tweak dynamics models and redeploy in minutes.
2
1
30
Did I get rid of hand-tuned parameters? Yes. Am I still hand-tuning more parameters? Also yes. I have a few knobs to address the sim-to-real gap -- mainly, the ESC/motor doesn't behave at all like the NN expects, need to bias and scale it a bit. Will learn that from data next.
1
2
4
Show this thread
I'm not going to spoil it, but (after charging the battery) this works way better than it has any right to. The car is now faster and more reliable than it's ever been. All my hand-engineered control system garbage is gone, completely rolled up into a 5k parameter neural net.
2
1
10
Show this thread
Decided to just see what happens if I run the sim-trained neural net on the car, with some safety rails around max throttle slew rates and such. Had to lie about velocity or it rolls the car over; otherwise it mostly just works.
3
5
63
Show this thread
I learned from this. This is my speed profile. Looks like I am too conservative on the right side of the track.
2
14
Dot color is speed; brighter is faster. Yeah, it has less room to explore in the tighter part, and the actor/critic algo learns to take a tight line. At first it can't even make it through there at all, and then the best line adjusts throughout training.
3
2
15
I'm gonna try to do proper offline reinforcement learning for and throw away all my manual parameter tuning for the next event. Just got a simple soft actor-critic implementation to work in sim as a baseline -- stochastic and deterministic policies pictured here.
6
4
44
Show this thread
DIY Robocars & Brazilian BBQ - Sat 10/1. Our track combines hairpin curves with an intersection for max danger. Take that, naïve AI and hacky line-followers! Only the smartest cars will survive. Spectators please register here: bit.ly/DIYrobocars
4
8
26
Had an great test today on track. However the car starts to drift at 40mph. Some experts recommended to change to soft tires. Will try this.
5
2
32
Always a good time working on my car at . Still got some work to do if I’m to beat however - he’s just too fast!
3
8
36
Show this thread
5
4
46
