A hox gene inspired generative approach to evolving robot morphology

Abstract

This paper proposes an approach to representing robot morphology and control, using a two-level description linked to two different physical axes of development. The bioinspired encoding produces robots with animal-like bilateral limbed morphology with co-evolved control parameters using a central pattern generator-based modular artificial neural network. Experiments are performed on optimizing a simple simulated locomotion problem, using multi-objective evolution with two secondary objectives. The results show that the representation is capable of producing a variety of viable designs even with a relatively restricted set of parameters and a very simple control system. Furthermore, the utility of a cumulative encoding over a non-cumulative approach is demonstrated. We also show that the representation is viable for real-life reproduction by automatically generating CAD files, 3D printing the limbs, and attaching offthe-shelf servo motors.

Some distance measures for morphological diversification in generative evolutionary robotics

“…The encoding used in the experiments is based on the one used in [22]. Inspired by how morphology is encoded in animals, the encoding is restricted to creating symmetric bodies with a spine-limb structure by describing a single limbgenerating program and a set of parameters specific to each spinal section.…”

“…As in [22], the limb-generating program is simply a variablelength list of genes, where each gene encodes the parameters used to describe the morphology of one segment. However, here a more generalized method is used to apply the section parameters during phenotype development.…”

A survey on evolutionary-aided design in robotics

“…The control side may include various neural network parameters or neural network wiring or other parameters unique to the specific controller type under use. In many of the applications the genome has a fixed length, with exceptions in [34,35].…”

“…Genotype representations in the literature use binary [36][37][38], integer [39], real number [40][41][42] and string [43] encoding with many applying either or a combination of the different encoding methods. The basic genome types could also be placed in a matrix [44], vector [34,[45][46][47], graph [48][49][50][51][52][53] and tree [54][55][56][57][58][59][60][61][62]. Further, in other representations, [34,63] applied the concept of a hox gene for encoding body morphology and [39,64] applied Compositional Pattern-Producing Networks (CPPN) based encoding .…”

“…The basic genome types could also be placed in a matrix [44], vector [34,[45][46][47], graph [48][49][50][51][52][53] and tree [54][55][56][57][58][59][60][61][62]. Further, in other representations, [34,63] applied the concept of a hox gene for encoding body morphology and [39,64] applied Compositional Pattern-Producing Networks (CPPN) based encoding . The CPPN encoding involves accepting arguments as inputs and generating a resultant graph which explains the connections between various functions.…”