In certain classes of environment, an embodied autonomous agent that can plan will outperform an equivalent agent that cannot. This talk will examine what is required for effective planning, and will explore the effects on planning systems of the constraints acting on evolved agents. Many of the apparently anomalous characteristics of consciousness will be shown to be consistent with the operations of a planning system in which a structure, process, or entity forming an Internal Agent Model (IAM) interacts with an internal world model to generate predictions about possible, desirable, and probable future actions. In particular, the IAM will be identified as the experiencing self. The implications of this approach for machine consciousness will be examined, and an attempt will be made to relate the theory to Aleksander’s axioms for consciousness, and to Sloman’s virtual machine architectures.