Title: Adapting the body of a robot to its real-world environment using artificial evolution
Speaker: Tonnes Nygaard, UiO
For robots to handle the numerous incidents that can affect them in the real world, they must be able to adapt to changes and unexpected events. The field of evolutionary robotics tries to solve some of these issues by automatically optimizing a robot for a specific environment. Most of the research in this field, however, uses only simplified representations of the robotic system in software simulations. Researchers working in physics simulators alone miss out on the natural noise, uncertainty, and richness of physical interaction in the real world that gives rise to new and exciting challenges and opportunities. Another tendency is to only optimize the control of the robot without considering the body. The theory of embodied cognition states that the mind, body, environment, and the interactions between them all serve as sources of cognition, and only optimizing the control of a robot does not offer the flexibility that arises from having an adaptable body as well.
This talk presents our work in real-world evolutionary optimization of both control and body on our custom four-legged robot, and we will discuss both challenges and solutions found encountered in our work.