Just so you know VF04 has been replaced with VF24. Also, I don't see any way of enforcing a constraint on the system with your beloved HSL.
Chalk one up for the GA. (Multiple design variables, multiple constraints, oh yeah!) It pays to be down in the trenches using this stuff on real problems, not academic trivia.
The entire car is far too large of a problem. Not that it couldn't be done, but in any engineering application it's always best to break a problem down into manageable pieces. A GA is like a smart-aleck child, if you make a mistake, it will exploit as much as it can to improve the solution. (Keeps you on your toes)
You might think this is a big money machine, but industry is reluctant to adopt designs produced by GAs. It's a great technology, but it hasn't been proven in a wide variety of applications. The auto industry for one is quick to point out the difference between prediction and simulation. The prior gives you approximate solutions while the later only gives you general trends
The search space is massive, but not so hilly that a GA can't function well.
Listen, a GA doesn't care about hills in the search space. This isn't a gradient based technique (hill climber). The strength of a GA is in it's ability to search a large design space and not get stuck on local maxima.
The beautiful thing about a G.A. is that you can use define as many design variables as you want(energy absorption, mass, manufacturability). If you can define it, you can optimize around it. We're currently using our software HEEDS (Hierarchical Evolutionary Engineering Design System) to conduct structural optimization of parts for the Auto industry. It's pretty wild when you can haver a piece of software actaully produce patents.
I first saw this idea at the U of M on their linux cluster in the Media Union.
They had all of their machines stacked in the original cases on three shelves that were on a rolling rack. It was a completely self contained unit. you can find these racks at www.globalindustrial.com This is what we are currently using for our clusters of PCs. Each cluster is mobile enough for moving since we're growing so fast. Don't get me wrong though, as soon as I can talk the higher ups into it we're going to sporting rack units, but for a $100 - $200 soultion per cluster, it's pretty nice (And they come in a variety of finishes black, chrome, stainless etc...)
The company I work for has a software package that can do this. The software can perform optimizations based on any type (and number) of design variables you can define. Currently we are using this technolosy to optimize the shape of structures for energy absorbtion in cars. It's pretty neat, the parts design themselves. However, we've also used this same technique to optimize the floor plans of manufacturing plants. Not to mention the uses for it in solving logistical problems. (For any of you optimization gurus out there, this is not a gradient based technique:)
All the previous ideas have been very good. I've been looking for something similar. I've also come across YACL (Yet Another Class Library) The library scources are free for download from http://www.cs.sc.edu/~sridhar/yacl.html
There's also a book on it Building Portable C++ Applications with YACL isbn: 0201832763
Slashdot Mirror
Chalk one up for the GA. (Multiple design variables, multiple constraints, oh yeah!) It pays to be down in the trenches using this stuff on real problems, not academic trivia.
nuff-z-nuff
Listen, a GA doesn't care about hills in the search space. This isn't a gradient based technique (hill climber). The strength of a GA is in it's ability to search a large design space and not get stuck on local maxima.
They had all of their machines stacked in the original cases on three shelves that were on a rolling rack. It was a completely self contained unit. you can find these racks at www.globalindustrial.com This is what we are currently using for our clusters of PCs. Each cluster is mobile enough for moving since we're growing so fast. Don't get me wrong though, as soon as I can talk the higher ups into it we're going to sporting rack units, but for a $100 - $200 soultion per cluster, it's pretty nice (And they come in a variety of finishes black, chrome, stainless etc...)
The company I work for has a software package that can do this. The software can perform optimizations based on any type (and number) of design variables you can define. Currently we are using this technolosy to optimize the shape of structures for energy absorbtion in cars. It's pretty neat, the parts design themselves. However, we've also used this same technique to optimize the floor plans of manufacturing plants. Not to mention the uses for it in solving logistical problems. (For any of you optimization gurus out there, this is not a gradient based technique :)
There's also a book on it Building Portable C++ Applications with YACL isbn: 0201832763
Just thought I'd throw that one into the pot too.
-C
This larger display will really add to the value of my I-opener. (At least perceived value... NOT!!)