🎓Learning to Embrace Change: Adaptive and Open-Ended Continual Learning for Embodied Autonomous Systems

Autonomous systems and robots operate in dynamic environments that change gradually or radically over time. To perform tasks in such ever-changing environments, the robot should continuously learn and adapt its capabilities. Continuous learning (CL) is a machine learning method that learns from continuous streams of data. CL aims to balance the trade-off between the capabilities to be maintained and those to be enhanced, a characteristic that current machine learning struggles with. The PhD thesis of MOHAMMAD WASIL focuses on using the physical interaction of a robot, such as exploiting spatio-temporal information about the environment, to obtain more robust data for addressing the continual learning problem.