Evolving Soft Robots in Tight Spaces

Abstract

Soft robots have become increasingly popular in recent years -and justifiably so. Their compliant structures and (theoretically) infinite degrees of freedom allow them to undertake tasks which would be impossible for their rigid body counterparts, such as conforming to uneven surfaces, efficiently distributing stress, and passing through small apertures. Previous work in the automated deign of soft robots has shown examples of these squishy creatures performing traditional robotic task like locomoting over flat ground. However, designing soft robots for traditional robotic tasks fails to fully utilize their unique advantages. In this work, we present the first example of a soft robot evolutionarily designed for reaching or squeezing through a small aperture -a task naturally suited to its type of morphology. We optimize these creatures with the CPPN-NEAT evolutionary algorithm, introducing a novel implementation of the algorithm which includes multi-objective optimization while retaining its speciation feature for diversity maintenance. We show that more compliant and deformable soft robots perform more effectively at this task than their less flexible counterparts. This work serves mainly as a proof of concept, but we hope that it helps to open the door for the better matching of tasks with appropriate morphologies in robotic design in the future. Figure 1: (counter-clockwise rotating-viewpoint time series starting from top left).An evolved soft robot reaches through a hole in the side of a cage surrounding it. The width of this aperture is smaller than any of the dimensions of this creature -thus a robot of the same size without a soft body would not be able to squeeze through it.

Material properties affect evolutions ability to exploit morphological computation in growing soft-bodied creatures

“…As these features are ubiquitous in natural systems, their presence here may suggest the potential for more competent and scalable virtual creatures. Moreover, these properties appear to be selected for by our task and environment, as this level of morphological regularity is not common in similar settings (Cheney et al, 2013(Cheney et al, , 2015Methenitis et al, 2015).…”

“…Third, differently from other works adopting VoxCad (Cheney et al, 2015(Cheney et al, , 2013Methenitis et al, 2015), in this experiment there is no fast-twitch actuation mechanism (i.e. the fast control based on an oscillating global signal is disabled).…”

“…Optimization. A multi-objective implementation of NEAT (Cheney et al, 2015) has been adopted. Before performing selection, pareto ranking is applied, according to three objectives: the order in which sorting is performed determines the relative importance of each of them.…”

Software toolkit for modeling, simulation, and control of soft robots

“…It is used in [55] to evaluate through simulation the walking capabilities of soft robots produced by genetic algorithms. In [56], the same authors added interaction between the robot and its environment. Nevertheless, with a voxel simulation, it is not possible to approximate some geometrical shapes without an exaggerated number of voxels which leads to an increased computation time.…”

Behavioral search drivers and the role of elitism in soft robotics

“…The authors proposed a developmental solution to the construction of more regular and symmetrical morphologies by sampling a compositional pattern producing network (CPPN) (Stanley, 2007) to construct forms. Subsequent work has moved towards a multi-objective approach, using Pareto optimization to incorporate complexity, energy usage, and age into guide the evolutionary process (Cheney et al, 2015;Kriegman et al, 2017). The results have produced seemingly lifelike gaits and behaviors 1 reminiscent of early work by Sims (1994).…”