Constructed Artificial Agents: Designing and Implementing Embodied Robots to Study Modularity in Evolution
Joshua Ridley ’17, Vassar College ’17 Jessica Ng, Vassar College ’16, John Loree, Vassar College ’16, Evan Altiero, Vassar College ’16 and Carl Bertsche, Joshua Bongard, Larry Doe, Ken Livingston, Nick Livingston, John Long, Jodi Schwarz, and Marc Smith
How evolvability itself evolved is an open question. To address this question scientifically, we use embodied robots to test the hypothesis that evolvability may evolve in concert with modularity. We designed Tadro14c, an autonomous swimming robot with OpenSCAD software. OpenSCAD allows one to construct a variety of three dimensional shapes including cylinders, spheres, cubes, and cones.
Cylindrical shapes of constrained but arbitrary lengths and radii were used to design each Tadro component. Cylinders are the building blocks for more complex shapes. They are an ideal starting point because they are easy to encode genetically and can be specified to undergo development by the use of promoters and transcription factors in genotype-to-phenotype (G→P) mappings. The design of the fin is vital in the propulsion and movement of the embodied robot. We used the Reynolds number, a dimensionless quantity used to predict flow patterns in varying fluid flow situations, along with the Bernoulli effect, the conservation of energy for flowing fluids, to determine the optimal fin shape needed. From these equations came a modified conventional airfoil design with a low camber, low drag, high speed, and thin wing section.
Components of the Tadro14c were manufactured using a MakerBot Replicator 5th generation 3D Printer. The printer prints using Polylactic Acid (PLA), a biodegradable plastic that allows for printing paper-thin layers and smooth-to-the-touch surfaces. In parallel, robot bodies of the same design were machined using acrylic materials (Anilam CNC milling machine). This will allow preliminary examination of the interaction of how the robot is embodied with how its control structure (see poster by Altiero, et al.) evolves.
Future work will include changing the morphology of the embodied robot through the placement of sensors and actuators within the specific parameters depicted by the G→P mapping system.