A Demonstration of Continuous Interaction with Elckerlyc

We discuss behavior planning in the style of the SAIBA framework for continuous (as opposed to turn-based) interaction. Such interaction requires the real-time application of minor shape or timing modifications of running behavior and anticipation of behavior of a (human) interaction partner. We discuss how behavior (re)planning and on-the-fly parameter modification fit into the current SAIBA framework, and what type of language or architecture extensions might be necessary. Our BML realizer Elckerlyc provides flexible mechanisms for both the specification and the execution of modifications to running behavior. We show how these mechanisms are used in a virtual trainer and two turn taking scenarios.