Developing an evolvable pattern generator using learning classifier systems

Classifying objects and patterns to certain categories is crucial for both humans and machines. Pattern classification has become an important topic in robotics research as it is applied in many scenarios (e.g. visual object detection in an autonomous robotics). Although autonomous learning of patterns by machines has advanced recently, it still requires humans to set-up the problem at an appropriate level for the learning technique. If the problem is too complex the system does not learn; conversely, too simple and the system does not reach its full potential performance level. In this work, a novel problem domain has been created that can be manipulated autonomously (i.e. scalable and evolvable patterns) to benefit autonomous systems. Experiments confirm that both the problem domain can be evolved and the problem solutions can be learnt lowering the requirement of human intervention in developing autonomous systems.