Efficient Learning for Autonomous Systems

Lead CI: Dirk Kroese

The movement, control, and planning of robotic and autonomous systems often involves statistical sampling on complex search spaces. For example, in Robotic Motion Planning the idea is to build up a “Probabilistic Road Map”, via Monte Carlo sampling on a high-dimensional space of parameters. This is often done on an ad hoc basis. The purpose of this proposal is to better understand how sampling and learning is best carried out for a variety of autonomic systems.