Cognitive control in cognitive robotics: Attentional executive control

摘要

We present an attentional control architecture for a robotic agent capable of adapting its deliberative and executive behavior to internal and external processes. In this framework, the agent is endowed with simple attentional mechanisms regulating the sensors sampling rates and action activations. The process of changing the frequency of sensors readings is interpreted as an increase or decrease of attention towards relevant tasks, activities, and processes. We propose a hybrid control architecture where these mechanisms are deployed at different levels of abstraction to regulate behavioral executions, execution monitoring, and dynamic planning. We also propose a benchmark scenario providing some empirical results showing the effectiveness of the attentional regulation.