SCHUMACHER, J; COMBIE, K; ENGEL, V; IRVING, K; LIVINGSTON, N; KOOB, T; LONG, J*; Vassar College; Vassar College; Vassar College; Vassar College; Vassar College; Shriners Hospital; Vassar College: Testing an Adaptation Hypothesis for Early Vertebrates with Biomimetic Robots
As adults, early fish-like vertebrates bore a notochord as their axial skeleton. Occluding the notochord, bony vertebral centra evolved independently and repeatedly in gnathostomes. While such convergence suggests a common selective environment, the details and complexities of plausible adaptive scenarios remain untested. We tested the hypothesis that centra were locomotor adaptations, increasing the body�s flexural stiffness and, hence, thrust generation and the individual�s ability to compete in a swimming-related task. Using autonomous, surface-swimming robots that mimic the phototactic behavior of ascidian tadpole larvae, we selected for improved navigation and station-keeping, rewarding, in a composite fitness function, swimming speed, path smoothness, quickness and proximity to the target. In each generation, we competed three robots, each differing only in the flexural stiffness and length of their tail. Tails were made with cylindrical axial skeletons of collagen hydrogel; stiffness was controlled by the amount of cross-linking. Based on relative fitness scores, we produced a new generation of tails using a genetic algorithm approach for mutation and mating. Both flexural stiffness and tail length were treated as quantitative, polygenic characters with a narrow-sense heritability of one. Over ten generations, we found a directional trend, caused by chance: tails became longer and less stiff without reducing overall navigational abilities. These results were contrary to our expectations. However, during generations with large differences in relative fitness, differential reproduction led to short-term improvements in speed, smoothness, quickness, and proximity, a result that supports the locomotor adaptation hypothesis. This project is supported by NSF grant DBI-0442269.
