Machine Consciousness and the Self Agent

The Self Agent (SA) is responsible for ISAC’s cognitive activities ranging from sensor signal monitoring to task monitoring to decision-making. For example, the SA integrates failure information from sensors and maintains information about the task-level status of the humanoid (see Section 5 for a detailed discussion). The SA cognitive aspects include recognition of human intention and selection of appropriate actions by forming plans and activating them.

Finally, we are developing an agent responsible for what might be termed executive control “to account for the control of information flow in memory tasks”. The structure and functions of the Self Agent are provided in Figure 1.

Figure 1. Structure/functions of Self Agent in IMA

In Figure 1, the Intention Agent determines the humanoid’s action based on the human’s intention, the state of the humanoid, and whether this action would conflict with the humanoid's current activities. If there is no conflict, the action can be performed, otherwise, the Intention Agent must resolve the conflict. Conflict resolution is currently handled based on the relative priority of the conflicting actions.

The Central Executive Controller (under construction) interacts with the Human Agent, the STM and the LTM to construct and invoke plans to perform various tasks. The goal of each generated plan is determined by input from the Intention Agent. Constructed plans are put into action by activating appropriate behaviors to form new procedural memory or motions, guided by the attention networks and sensory information received from the SES.