Final Deadline
My internship at the University of Tsukuba is not over but I have finished my final presentation and final report on the Evolution of Voxel-Soft-Robots.
Here is the report and here is the final presentation.
The videos of the robots are accessible through the Results section of the report.
What I Found
- A single genome can encode two bodies and two controllers, one pair per environment, well enough to reach good fitness on both — and more importantly, the bodies and controllers it produces are coherent with the environment they face, not just scoring well. But ploidy alone answers nothing. I crossed two ploidies with two levels of phenotypic plasticity — developmental where an environmental cue enters the body ANN, and also a chromosome-level where the cue decides which chromosome is expressed — and the result changes between them. Under developmental plasticity, diploidy is a source of noise : every allele keeps moving at once, so nothing ever holds still long enough for the rest to settle around it. Under chromosome-level plasticity, the results changes — diploidy is the only representation that adapts to both tasks, while the haploid one locks into a single one.
So basically, I started with the question — Does diploidy help ? But maybe the right question is — Under what conditions does diploidy help ?
Acknowledgements
Again, all my thanks to my supervisor Pr. Claus Aranha for welcoming me in his lab and for his invaluable support, guidance and patience, and also my thanks to Pr. Dennis Wilson for his mentorship before and during this project.