PhD Defence | Where is the robot?
Karine Miras’s thesis explores how robots can evolve like living beings, i.e., artificial life. Can their whole bodies and brains be subject to evolution (natural selection), i.e., they are born, mate, have children, and the evolutionary loop keeps on going like in nature. Miras completed her research under the supervision of Guszti Eiben and Evert Haasdijk (both from VU Amsterdam).
How can you design a robot that presses a button on an unknown distant planet if you do not even know what the button looks like? Answer: You do not design this robot, but a system that is able to auto design robots through evolution can! Most of the research is in simulation, but in this research, some of the robots are actually printed, and the long term vision is making the whole process feasible to run solely in real hardware.
During Miras’s research, one of the objectives was creating an artificial (robot) DNA structure that allows itself to be regulated by the environment. Miras answers multiple questions related to how the environment influences robot artificial life in two ways:
- Indirect influence: the environment creates a selection pressure for determined phenotypic and behavioral traits to emerge; while in nature the influence of the environment in shaping phenotypic traits is a well-known “obvious” notion, for artificial life systems is it much harder to achieve the same results. This happens because in nature there is a multitude of factors that cannot be recreated in the simplistic current experimental setups used. Still, Miras managed to do that, i.e., demonstrates cases when a change in the environment induces a change in phenotypic and behavioral traits!
- Direct influence: the environment regulates gene expression allowing phenotypic plasticity (bodies and brains changing according to changes in the environmental conditions); Miras proved that just like in nature, we can reproduce such processes and obtain benefits to the robots being evolved (better, more adaptable robots!).
Finally, Miras found one very curious thing: during evolution, she observed the emergence of robots that very much locomote like animals and even have a similar shape to them. Note that the robots were NOT directly designed to look like animals, but these solutions simply emerged from the artificial system, once it is based on the same principles that allow animals and humans to emerge. This proves the potential of the system to create intelligence, not only through the mind (brain) but through the body!