Evolution of Adaptive Behaviour

its too bad we could not try doing some stuff like this with the reprap [www.plosbiology.org]

it is about Evolution of Adaptive Behaviour in Robots by Means of Darwinian Selection

Edited 1 time(s). Last edit at 01/30/2010 04:06PM by dissidence.

---

[mike-mack.blogspot.com]

Thats true...since we have the new evolution of Robots

It'd be interesting to apply a better system for the evolution of RepRap...

Some sort of scoring system for print speed, print accuracy, print success rate, power consumption, print size; A way of relating these scores to the "genetics" of the machine (by recording initial frame, extruder, electronics), the success of alterations (by recording the improvement/reduction of each score for each modification, and the number of people who adopted the modification).

A forum user wanting to build a new printer could look at the current score card for each basic printer design, and the relative score card for each design modification. The user then builds based on their criteria (faster builds, or more accurate builds, or bigger builds), and records the "genome" of their printer for statistical purposes, as well as the Score Card for the final printer.

This would result on the more successful designs and modifications being adopted more easily by new users, and thus getting a better rating, instead of just adopting the Status Quo, or not being sure which of the many design variations are better, leading to a better breed of printers for all.

No worries about awesome modifications being overlooked in favour of more "standard" designs, because by adding a scorecard for the design, showing how much better your printer performed with it, any user searching for a [X] or wanting to replace that wobbly [Y] would do a search by component type and sort by highest improvement score, and see that awesome mod.

Anyone got a website/database that could be used to set up a system like this?

This sort of happens anyhow by proxy. The better and faster machines get more attention and consequently more people build and develop them.

Unfortunately this Darwinian evolution leads to dead ends. Its been tried many times, the examples I can think of are for aircraft wing designs and for wind turbine development. These make subtle changes and improved things, but no radical alternatives came from them.

The advantage us humans have over evolution is that we can think outside the box. We can chose to evolve our designs or come up with something completely new. Points based "top of the pops" charts will just channel people into rehashing the same design. OK more people will get involved but most of them will be sheep.

Take for example the delta robots that are popping up all over the place. If we just used pure evolutionary methods like you are suggesting the delta robot would not have evolved from the cartesian prusa.

We probably could, the robots they made were printed with a FDM type printer, and they have the software they used to evolve robots available for download here: [www.demo.cs.brandeis.edu]

It shouldn't be too hard of a task to write these evolved robots into STL files using openscad and what not.


Though if you are referring to evolving printers that's going to be a bit harder.

At the very least we could do some evolutionary structural optimization to find the stiffest lightest frame for reprap. If we can figure out the loading scenarios for reprap, this should be a very doable task.

I was reading the content of your link and its interesting to read but visioning on doing that is difficcult.

... I participated in the Golem project, but found the results a bit disappointing - most results converged to the typical triangular 'gliders' ... could be the limited design rules

More interesting was the Framsticks-project: [www.framsticks.com]

---

Viktor
--------
Aufruf zum Projekt "Müll-freie Meere" - [reprap.org] -- Deutsche Facebook-Gruppe - [www.facebook.com]

Call for the project "garbage-free seas" - [reprap.org]

This is cool, I like this topic.

I am a developer for DarwinBots, and I suggest you take it once step further:

You can get a secluded area with a bunch of metal and plastic materials. Then you can instruct the printer to build 'ants' that go around collecting these materials for the 3D printer. Then you can instruct the printer to print more 'ants' or print a new 'queen' i.e. a new 3D printer. You can add copy errors etc. to the blue print as you go. It will be interesting to see what will result because we are building actual artificial life, it is no longer just a computer simulation.

I hope I am scaring too much people with this. And you guys should defiantly have the resources to try it. :p