Loading ...
Sorry, an error occurred while loading the content.

6238Re: [neat] RE: Environmental Influence on the Evolution of Morphological Complexity in Machines

Expand Messages
  • Joshua Auerbach
    Jan 7, 2014
    • 0 Attachment
      Thanks Ken and Jeff!

      Sorry for the delayed response, but I've been busy traveling back to Europe. 

      I agree that there is potential for more progress in this and other similar directions (as Jeff's work with Nick Cheney and Hod Lipson has also demonstrated), but I agree that the computational cost of running these simulations really is the bottleneck at the moment, and why so far we haven't had much luck complexifying the environments in other ways that may be more reminiscent of natural surfaces.  I did a little bit more work towards allowing the CPPNs to decide their joints and degrees of freedom in my ALife paper from 2012: http://www.cs.uvm.edu/~jeauerba/publications/auerbach_bongard_alife_2012.pdf
      I think there is a lot of value in finding appropriate ways to do that, but I returned to this simpler system here in order to focus on the morphological complexity.

      I do plan on continuing this work, but am also involved in a new project here at EPFL at the moment, so have had to put this stuff on the back burner for the time being.  I do, at some point, hope to find a way to integrate it with the other things I am working on here.  Without giving away too much of future plans I will say that am excited to have access to some really high quality additive manufacturing facilities and hope to be bringing these or other similarly complex morphologies into the real world.

      Best regards,

      On 01/07/2014 04:43 AM, kstanley@... wrote:

      Hi Josh, yes congratulations, I agree with Jeff that it's nice to see this work in PLoS Comp. Bio.  As you know I really like the unique way you use CPPNs to represent these bodies.  Their nuanced shapes are different from the conventional "connected blobs" type of virtual creature and to me feel more natural, so I feel like there's potential for more progress in this direction, though I understand simulating these creatures is pretty expensive.  Do you have any plans to develop more creatures of this type, maybe giving their CPPNs somehow the ability to decide their own joints and degrees of freedom?  The arrow of complexity is also a really interesting topic and it's nice to see it elevated in this kind of journal. 

      Dr. Joshua E. Auerbach
      Postdoctoral Researcher
      Laboratory of Intelligent Systems
      École Polytechnique Fédérale de Lausanne
    • Show all 4 messages in this topic