Genetic redundancy in evolving populations of simulated robots

Area of Science:

• Evolutionary computation
• Artificial life
• Robotics

Background:

• Biological systems often exhibit genetic redundancy, which some researchers propose enhances evolvability.
• The role and impact of redundancy in evolutionary processes remain an active area of investigation.

Purpose of the Study:

• To experimentally investigate the hypothesis that genetic redundancy contributes to evolvability.
• To explore the effects of genetic redundancy on the evolutionary dynamics of simulated robot populations.

Main Methods:

• Utilized a genetic algorithm to evolve simulated robots controlled by artificial neural networks.
• Manipulated genetic redundancy by varying genotype size and measured it using systematic lesioning.
• Assessed population fitness, evolvability, and diversity.

Main Results:

• Populations with larger, redundant genotypes achieved systematically higher fitness compared to those with smaller genotypes.
• Despite smaller genotypes having sufficient computational power for optimal fitness, redundant populations exhibited greater evolvability and diversity.
• Enhanced evolvability in redundant populations is attributed to the exploration of extensive neutral networks in genotype space.

Conclusions:

• Genetic redundancy significantly enhances the evolvability and diversity of evolving robot populations.
• Redundancy facilitates the exploration of neutral networks, a key mechanism for evolutionary innovation.
• The findings support the conjecture that redundancy in functional genomic components can be advantageous for evolution in both artificial and biological systems.