Integrating Planning and Recognition to Close the Interaction Loop

In many real-world domains, the presence of machines is becoming more ubiquitous to the point that they are usually more than simple automation tools for one-way interaction. As part of the environment amongst human users, it is necessary for these computers and robots to be able to interact back reasonably by either working independently around them or participating in a task, especially one with which a person needs help. Such interactions are now everywhere ranging from robots around homes and factories to virtual agents in mobile devices, video games, and automated dialogue systems. While interactive robots and computer systems have been implemented for various domains, most are specifically designed for a given domain such as industrial robotics (Levine and Williams 2014; Wurman, D’Andrea, and Mountz 2007), elderly care (Schwenk, Vaquero, and Nejat 2014; Fasola and Matarić 2013), etc. Just as there are domain-independent heuristics that can effectively find optimal solutions for any classical planning problem (Hoffmann and Nebel 2001; Helmert 2006), I introduce a domainindependent approach to performing interaction based on the integration of several research areas in artificial intelligence, particularly planning and plan recognition. This interactive procedure requires several steps performed indefinitely as a loop: recognizing the user and environment from sensor data, interpreting the user’s activity and motives, determining a responsive behavior, beginning to perform the behavior, and then recognizing everything again to confirm the behavior choice and replan if necessary. At the moment, the research areas addressing these steps, activity recognition, plan recognition, intent recognition, and planning, have all been primarily studied independently. However, pipelining each independent process can be risky in real-time situations where there may be enough time to only run a few steps. This leads to a critical question: how do we perform everything under time constraints? In this thesis summary, I propose a framework that integrates these processes by taking advantage of features shared between them. This includes my current work towards this preliminary system and outlines how I plan to complete the integration for a time-constrained interaction loop.