Idiotypic Networks for Evolutionary Controllers in Virtual Creatures

Abstract

We propose a novel method for evolving adaptive locomotive strategies for virtual limbless creatures that addresses both functional and non-functional requirements, respectively the ability to avoid obstacles and to minimise spent energy. We describe an approach inspired by artificial immune systems, based on a dual-layer idiotypic network that results in a completely decentralised controller. Starting from a system initialised with five non-adaptive locomotion strategies, we show that an adaptive controller can evolve that both minimises energy requirements and maximises distance covered when compared to the initial strategies.

Artificial Immunology for Collective Adaptive Systems Design and Implementation

“…Results show that the approach is able to adapt the initial movements into new strategies, which for example enable it to avoid simple obstacles and additionally minimise energy consumption, compared to the initial strategies. For a more detailed analysis of these assertions, the reader in once again referred to [Capodieci et al 2014b]. As in the previous study, the ability to exhibit self-expression provides improved performance with respect to each individual coordination pattern.…”

“…Rather than focussing on results, this section highlights the relevant design choices in terms of the framework and discusses the implications of these choices. As each study has been described in detail in previously published conference proceedings, the reader is referred to the corresponding publication 2014b;2014a] for a more detailed discussion of the results.…”

“…their tuple values are mathematically averaged) to create a new antibody. A second mutation operator is activated to create new antibodies from a single antibody whenever the first mutation operator fails to deliver expected results (described in detail in [Capodieci et al 2014b].…”