Motivated Goal and Action Selection

A large part of current research into autonomous control is concerned with building agents that can perform useful functions in real-world domains. Consequently, the behaviour of the system must be exible; the behaviour of an agent must reeect changes in context. This position paper is concerned with the use of motivation as a control mechanism in goal and action selection within an artiicial system. It is argued that existing motivated behaviour-based control systems do not eeectively give suucient exibility; and that a diierent system based on explicit goal and action representations can more eeectively control an autonomous system in a real-world domain.