Evolutionary Developmental Robotics: Improving Morphology and Control of Physical Robots.

Area of Science:

• Robotics and Evolutionary Computation
• Developmental Biology in Artificial Systems

Background:

• Evolutionary algorithms are used for robot design, but complex design spaces hinder efficient solution discovery.
• Previous methods often lack the ability to simulate developmental processes in physical robots.

Purpose of the Study:

• To introduce and evaluate an evolutionary-developmental (evo-devo) approach for real-world robot design.
• To investigate the impact of ontogenetic morphological changes (leg growth) on robot performance.
• To explore the potential of developmental stages in improving robotic systems.

Main Methods:

• Implementing an evo-devo experiment with physical robots capable of altering leg size during their 'lifetime'.
• Generating diverse robot leg morphologies using evolutionary processes and additive fabrication.
• Conducting experiments comparing evo-devo approaches with pure evolutionary methods.

Main Results:

• Demonstrated the first physical robot experiment incorporating ontogenetic morphological changes (leg growth).
• Showcased improved robot performance through the addition of a developmental stage.
• Identified nonlinear system-environment interactions leading to complex locomotion patterns.

Conclusions:

• The integration of developmental stages in robotics can significantly enhance robot performance and adaptability.
• Evo-devo principles offer a promising avenue for creating more sophisticated and resilient physical robots.
• Future robots will evolve and grow, interacting dynamically with their environments.