It seems to always get into a rut where one design lucks out and dominates generation after generation, with no mutations producing anything even close to working. Like, the top ten don't change after hundreds of generations. Maybe this is just an attribute of genetic algorithms. They quickly zero in on something kind of good, and then get stuck at this local maxima. Or maybe I need to just play around with the Mutation Rate and Mutation Size settings.
Watch cars evolve using genetic algorithm
31–40 of 76 posts
Re: Watch cars evolve using genetic algorithm
#32Considering the 2D nature of the simulation, aren't they more motorbikes than cars?
Re: Watch cars evolve using genetic algorithm
#33It seems to always get into a rut where one design lucks out and dominates generation after generation, with no mutations producing anything even close to working. Like, the top ten don't change after hundreds of generations. Maybe this is just an attribute of genetic algorithms. They quickly zero in on something kind of good, and then get stuck at this local maxima. Or maybe I need to just play around with the Mutat…
Sounds like you're running with a single elite clone. That's a really bad idea in genetic heuristics for exactly the phenomenon you mention.
Re: Watch cars evolve using genetic algorithm
#34I wrote the original version of this nearly 20 years ago. Cool to see it still pop up here from time to time. Still runs in the browser thanks to Ruffle: https://peteshadbolt.co.uk/posts/ga/
I remember running this in the background any time I was in the computer lab at school. Thanks for making this back in the day :).
There were a few genetic algorithm "polygons approximate picture" pages back in that era as well.
Re: Watch cars evolve using genetic algorithm
#35Interesting. Is there a way to do this in a 3d physics based simulation environment. It would be cool to see if a genetic algorithm could be used to discover new aerodynamic configurations for drones/other platforms in simulation. I don't know enough about genetic algorithms to say for certain. Anyone have any reference materials for someone that's just started looking into this?
Not exactly what you're asking but Topology Optimization is now a standard feature of the big CAD packages. It allows the designer to express various constraints and goals, then a combination of gradient methods and genetical algorithms are used to find an optimized part. Example: https://www.solidworks.com/media/topology-optimization
Re: Watch cars evolve using genetic algorithm
#36I wrote the original version of this nearly 20 years ago. Cool to see it still pop up here from time to time. Still runs in the browser thanks to Ruffle: https://peteshadbolt.co.uk/posts/ga/
I worked on Adobe’s Flash Player team at that time and I remember using your Flash movie as a test case. :)
Re: Watch cars evolve using genetic algorithm
#37Funny how the landscape never changes, even though the generations come and go.
Re: Watch cars evolve using genetic algorithm
#38Slight bug: there's no road past about 280m, all cars fall into an endless pit
Re: Watch cars evolve using genetic algorithm
#39Slight bug: there's no road past about 280m, all cars fall into an endless pit
Time comes for us all
Re: Watch cars evolve using genetic algorithm
#40It seems to always get into a rut where one design lucks out and dominates generation after generation, with no mutations producing anything even close to working. Like, the top ten don't change after hundreds of generations. Maybe this is just an attribute of genetic algorithms. They quickly zero in on something kind of good, and then get stuck at this local maxima. Or maybe I need to just play around with the Mutat…
>Maybe this is just an attribute of genetic algorithms
It happened to crocodiles and it could happen to you too.