Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Using neuroevolution for designing soft medical devices

Hugo Alcaraz-Herrera, Michail-Antisthenis Tsompanas, Andrew Adamatzky, Igor Balaz

TL;DR

This study automates the design of soft catheter actuators using neuroevolution (AFPO, NEAT, HyperNEAT) evaluated in a voxel-based 3D simulator. By representing actuator morphologies with indirect encodings via neural networks and optimizing for maximum bending, the work compares how topology, recombination, and geometric substrates influence performance and robustness. Results show NE-based methods generally yield more robust morphologies than AFPO, with NEAT often delivering the best balance between bending capability and voxel count, while HyperNEAT's performance depends on substrate design. The findings suggest neuroevolution can provide test-bed morphologies for developing specialized controllers and highlight avenues for manufacturing-aware filtering and advanced substrates.

Abstract

Soft robots can exhibit better performance in specific tasks compared to conventional robots, particularly in healthcare-related tasks. However, the field of soft robotics is still young, and designing them often involves mimicking natural organisms or relying heavily on human experts' creativity. A formal automated design process is required. We propose the use of neuroevolution-based algorithms to automatically design initial sketches of soft actuators that can enable the movement of future medical devices, such as drug-delivering catheters. The actuator morphologies discovered by algorithms like Age-Fitness Pareto Optimization, NeuroEvolution of Augmenting Topologies (NEAT), and Hypercube-based NEAT (HyperNEAT) were compared based on the maximum displacement reached and their robustness against various control methods. Analyzing the results granted the insight that neuroevolution-based algorithms produce better-performing and more robust actuators under different control methods. Moreover, the best-performing morphologies were discovered by the NEAT algorithm. As a future work aspect, we propose using the morphologies discovered here as test beds to optimize specialized controllers, enabling more effective functionality towards the desired deflections of the suggested soft catheters.

Using neuroevolution for designing soft medical devices

TL;DR

This study automates the design of soft catheter actuators using neuroevolution (AFPO, NEAT, HyperNEAT) evaluated in a voxel-based 3D simulator. By representing actuator morphologies with indirect encodings via neural networks and optimizing for maximum bending, the work compares how topology, recombination, and geometric substrates influence performance and robustness. Results show NE-based methods generally yield more robust morphologies than AFPO, with NEAT often delivering the best balance between bending capability and voxel count, while HyperNEAT's performance depends on substrate design. The findings suggest neuroevolution can provide test-bed morphologies for developing specialized controllers and highlight avenues for manufacturing-aware filtering and advanced substrates.

Abstract

Soft robots can exhibit better performance in specific tasks compared to conventional robots, particularly in healthcare-related tasks. However, the field of soft robotics is still young, and designing them often involves mimicking natural organisms or relying heavily on human experts' creativity. A formal automated design process is required. We propose the use of neuroevolution-based algorithms to automatically design initial sketches of soft actuators that can enable the movement of future medical devices, such as drug-delivering catheters. The actuator morphologies discovered by algorithms like Age-Fitness Pareto Optimization, NeuroEvolution of Augmenting Topologies (NEAT), and Hypercube-based NEAT (HyperNEAT) were compared based on the maximum displacement reached and their robustness against various control methods. Analyzing the results granted the insight that neuroevolution-based algorithms produce better-performing and more robust actuators under different control methods. Moreover, the best-performing morphologies were discovered by the NEAT algorithm. As a future work aspect, we propose using the morphologies discovered here as test beds to optimize specialized controllers, enabling more effective functionality towards the desired deflections of the suggested soft catheters.
Paper Structure (18 sections, 11 equations, 10 figures, 2 tables)

This paper contains 18 sections, 11 equations, 10 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Methodology
  4. AFPO
  5. NEAT
  6. HyperNEAT
  7. Experimental setup and Results
  8. Experimental setup
  9. NEAT configuration
  10. HyperNEAT configuration
  11. Analysis of the performance
  12. Finding the fittest actuator
  13. Exploring the robustness of the fittest actuator
  14. Identifying the minimum volume of the fittest actuators
  15. Error analysis
  16. ...and 3 more sections

Figures (10)

  • Figure 1: An example of a population moving in a two-dimensional Pareto space where the vertical axis is fitness and the horizontal axis is age. Figure adapted from Schmidt2010.
  • Figure 2: Example of a genotype and its phenotype under NEAT. Figure adapted from Stanley2002efficient.
  • Figure 3: Two typical configurations of substrates utilised in HyperNEAT: (a) two-dimensional, and (b) three-dimensional.
  • Figure 4: Substrate employed under HyperNEAT for actuator generator.
  • Figure 5: Mean performance of AFPO, NEAT, and HyperNEAT showing $\pm$95% confidence interval (shaded region).
  • ...and 5 more figures