Abstract
An ambitious goal in evolutionary robotics is to evolve increasingly complex robotic behaviors with minimal human design effort. Reaching this goal requires evolutionary algorithms that can unlock from genetic encodings their latent potential for evolvability. One issue clouding this goal is conceptual confusion about evolvability, which often obscures the aspects of evolvability that are important or desirable. The danger from such confusion is that it may establish unrealistic goals for evolvability that prove unproductive in practice. An important issue separate from conceptual confusion is the common misalignment between selection and evolvability in evolutionary robotics. While more expressive encodings can represent higher-level adaptations (e.g. sexual reproduction or developmental systems) that increase long-term evolutionary potential (i.e. evolvability), realizing such potential requires gradients of fitness and evolvability to align. In other words, selection is often a critical factor limiting increasing evolvability. Thus, drawing from a series of recent papers, this article seeks to both (1) clarify and focus the ways in which the term evolvability is used within artificial evolution, and (2) argue for the importance of one type of selection, i.e. divergent selection, for enabling evolvability. The main argument is that there is a fundamental connection between divergent selection and evolvability (on both the individual and population level) that does not hold for typical goal-oriented selection. The conclusion is that selection pressure plays a critical role in realizing the potential for evolvability, and that divergent selection in particular provides a principled mechanism for encouraging evolvability in artificial evolution.
Highlights
Natural evolution is an unguided process that has produced organisms with functionalities far exceeding the products of current human engineering
Note that while this paper focuses on evolutionary robotics (ER), the insights likely generalize to many applications of evolutionary algorithms (EAs) beyond the black box setting (Doncieux and Mouret, 2014); for this reason, the term EA assumes domains in which observable behavior results from an evaluation
This paper focuses on this latter factor, because while there exist many interesting proposals for more expressive encodings (Spector and Robinson, 2002; Risi et al, 2010; Lehman and Stanley, 2011b), in practice realizing their potential for evolvability remains difficult (Edmonds, 2001; Spector and Robinson, 2002; Clune et al, 2008)
Summary
Natural evolution is an unguided process that has produced organisms with functionalities far exceeding the products of current human engineering. Biological evolution is well studied, the abstract mechanisms through which cascades of increasingly complex functionalities evolve are not deeply understood Supporting this claim, artificial evolutionary processes so far cannot reproduce biological levels of behavioral complexity. One particular source of confusion addressed here concerns the level of organization (e.g. the level of the individual or the population) in which it is most important to consider and encourage evolvability. To address such confusion, this paper aggregates from an ongoing research agenda (Lehman and Stanley, 2011b; Wilder and Stanley, 2015) one coherent vision of evolvability. To create more prolific ER algorithms may require increasing the potential for evolvability in encodings, focusing on populationlevel evolvability instead of on individual-level evolvability, and guiding search through divergent selection
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
