Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Unconventional Hexacopters via Evolution and Learning: Performance Gains and New Insights

Jed Muff, Keiichi Ito, Elijah H. W. Ang, Karine Miras, A. E. Eiben

TL;DR

This work investigates evolvable hexacopter morphologies paired with learnable controllers to surpass traditional designs on navigation tasks. By decoupling morphology evolution from controller learning under the Triangle of Life, the authors demonstrate that non-conventional, irregular drones can achieve substantially higher performance across Circle, Slalom, Shuttle Run, and Figure8 tasks. They introduce domain-agnostic metrics for learning dynamics and symmetry analyses, revealing that evolved morphologies facilitate faster, sometimes more volatile learning, signaling a broader principle that morphology biases learnability in embodied AI. The study provides both practical drone-design insights and a methodological toolbox for rigorously analyzing the interplay between evolution and learning in complex embodied systems.

Abstract

Evolution and learning have historically been interrelated topics, and their interplay is attracting increased interest lately. The emerging new factor in this trend is morphological evolution, the evolution of physical forms within embodied AI systems such as robots. In this study, we investigate a system of hexacopter-type drones with evolvable morphologies and learnable controllers and make contributions to two fields. For aerial robotics, we demonstrate that the combination of evolution and learning can deliver non-conventional drones that significantly outperform the traditional hexacopter on several tasks that are more complex than previously considered in the literature. For the field of Evolutionary Computing, we introduce novel metrics and perform new analyses into the interaction of morphological evolution and learning, uncovering hitherto unidentified effects. Our analysis tools are domain-agnostic, making a methodological contribution towards building solid foundations for embodied AI systems that integrate evolution and learning.

Unconventional Hexacopters via Evolution and Learning: Performance Gains and New Insights

TL;DR

This work investigates evolvable hexacopter morphologies paired with learnable controllers to surpass traditional designs on navigation tasks. By decoupling morphology evolution from controller learning under the Triangle of Life, the authors demonstrate that non-conventional, irregular drones can achieve substantially higher performance across Circle, Slalom, Shuttle Run, and Figure8 tasks. They introduce domain-agnostic metrics for learning dynamics and symmetry analyses, revealing that evolved morphologies facilitate faster, sometimes more volatile learning, signaling a broader principle that morphology biases learnability in embodied AI. The study provides both practical drone-design insights and a methodological toolbox for rigorously analyzing the interplay between evolution and learning in complex embodied systems.

Abstract

Evolution and learning have historically been interrelated topics, and their interplay is attracting increased interest lately. The emerging new factor in this trend is morphological evolution, the evolution of physical forms within embodied AI systems such as robots. In this study, we investigate a system of hexacopter-type drones with evolvable morphologies and learnable controllers and make contributions to two fields. For aerial robotics, we demonstrate that the combination of evolution and learning can deliver non-conventional drones that significantly outperform the traditional hexacopter on several tasks that are more complex than previously considered in the literature. For the field of Evolutionary Computing, we introduce novel metrics and perform new analyses into the interaction of morphological evolution and learning, uncovering hitherto unidentified effects. Our analysis tools are domain-agnostic, making a methodological contribution towards building solid foundations for embodied AI systems that integrate evolution and learning.

Paper Structure

This paper contains 21 sections, 15 equations, 10 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem description: drones and tasks
  4. Design space: hexacopcter drones
  5. Tasks
  6. Evolution
  7. Initial Population
  8. Representation
  9. Mutation Operator
  10. Repair Operator: Ensuring Rotor Separation
  11. Fitness function
  12. Reinforcement Learning
  13. Experiment setup
  14. Human versus evolved designs
  15. Task performance of evolved designs
  16. ...and 6 more sections

Figures (10)

  • Figure 1: Illustration of the Triangle of Life framework, cf. Eiben2013, for evolutionary robotics underpinning the setup used in this paper.
  • Figure 2: Hexacopter type drone to illustrate the design space of morphologies.
  • Figure 3: Diagrammatic representation of the tasks (using a quadrocopter for illustration).
  • Figure 4: Schematic image of a drone with three arms and propellers to illustrate the effect of the parameters that represent the body layout.
  • Figure 5:
  • ...and 5 more figures