Humans are highly skilled at abstraction reasoning, i.e. the ability to abstract patterns in data and reapply them to similar/related states. Given a few examples of an ‘office chair’, we can recognise novel examples of office chairs without retraining. Unfortunately, mobile robots, such as autonomous cars, look for regular patterns rather than abstract patterns. Hence autonomous vehicles have a tendency to collide with static emergency vehicles as although the underlying pattern is constant the observed pattern is affected by lighting conditions, variations in design and unusual poses of the vehicle. This doctorate research will consider symbolic learning for abstraction, initially using the abstraction reasoning corpus. Once promising techniques have been identified, these will be transferred to a mobile robot to identify common patterns within an environment.