The University of Oslo is closed. The PhD defence and trial lecture will therefore be fully digital and streamed directly using Zoom. The host of the session will moderate the technicalities while the chair of the defence will moderate the disputation.
Ex auditorio questions: the chair of the defence will invite the audience to ask ex auditorio questions either written or oral. This can be requested by clicking 'Participants -> Raise hand'.
Title: "Are evolutionary algorithms good alternatives to (deep) reinforcement learning algorithms in robotics?"
Main research findings
Future robotic platforms and tasks require optimization algorithms that are capable of adapting both the control system as well as the body of the robot itself. To achieve this, optimization algorithms that are capable of discovering high performing solutions are needed. One important property of such optimization algorithms is the capacity to discover diverse solutions that represent different approaches to solving the problem. By promoting diversity among solutions, new and interesting approaches may be discovered - creating stepping stones to further the optimization process.
The goal of this thesis is to explore how diversity-promoting optimization algorithms can be used to overcome some of the fundamental challenges in robotics. By combining existing techniques from the field with an innovative optimization algorithm - called MAP-Elites, this thesis generates new insight into the optimization process. Through optimizing the control system of a quadruped robot as well as evolving modular robots, the result of this thesis shows that a diversity of solutions could hold the key to overcoming some of the fundamental challenges in the field of Evolutionary Robotics.
Contact information to Department: Anniken R. Birkelund