Evolving Modular Robots: Challenges and Opportunities

Abstract

The body of robots and their controllers need to be adapted to the task that they carry out. While it is possible to design and optimize free-form morphologies, its physical implementation consumes too many resources. In contrast, modular robots provide a feasible approach to design robotic morphologies that can be deployed in minutes, making them a suitable tool to implement virtual creatures. In this article, we tackle the main challenges to consider when evolving modular robots and mention some opportunities that these systems can provide.

Evolving Modular Soft Robots without Explicit Inter-Module Communication using Local Self-Attention

“…Modularity allows robotic systems to present various kinematic configurations beyond what a fixed architecture can, and such robots are usually optimal to solve many robotic tasks [16,63]. Modularity in robotic systems takes the form of fabricating physical parts that are interchangeable for a single robot, or designing independent robots that participate in a common task adaptively [18]. Due to the flexibility, versatility, and robustness to changing environmental conditions [78], we are witnessing an increasing number of works on modular robots [31], including soft ones [67].…”

Evolving Hebbian Learning Rules in Voxel-Based Soft Robots

“…Thanks to their intrinsic modularity, they have emerged as a relevant formalism for designing state-of-the-art soft robotics systems [22]. Modularity is a desirable property for robotic systems [23], as it makes the morphological search space compact without harming expressivity; indeed, interesting processes (e.g., development [24]) are still plausible.…”