Antony Waldock Title : Learning through Uncertainty Abstract: The construction of an intelligent controller for a mobile robot is a time-consuming and error prone process. Existing methods rely on the construction of accurate analytical models of the environment or manual decomposition of the controller into manageable sub problems. Learning techniques provide attractive advantages over manual engineering but remain un-interpretable and highly domain dependant. This talk will outline some initial experiments into a novel learning technique capable of constructing an interpretable controller using the amount of uncertainty experienced.