Bringing Up Robot: Fundamental Mechanisms for Creating a Self-Motivated, Self-Organizing Architecture

A signi cant pitfall exists in task-oriented robot design|an inherent anthropomorphic bias. Traditional research in the design of robots has attempted to get robots to do the tasks a human can, and to do them in the way a human would. The pitfall in this approach is that our human conceptualization of how to solve a particular task is rmly grounded in our unique sensory experiences and motor skills. A robot that is equipped with very dierent sensor and motor capabilities cannot easily share our conceptualizations. Instead, it is more fruitful to eliminate the anthropomorphic bias by adopting a developmental approach. This paper proposes a multi-level, cascaded discovery and control architecture for developmental robotics. The key components of this developmental architecture are mechanisms for abstraction and anticipation in the context of a model of self-motivation.