This paper presents work on reproducing complex forms of animal learning in simulated Khepera robots using a behaviour-based approach. The work differs from existing behaviour-based approaches by implementing a path of hypothetical evolutionary steps rather than using automated evolutionary development techniques or directly implementing sophisticated learning. Following a step-wise approach has made us realise the importance of maximising the number of behaviours and activities included on one level of complexity before progressing to more sophisticated solutions. We call this inclusion behavioural holism and argue that successful approaches to complex behaviour based robotics must be both step-wise and holistic.
展开▼
