Evaluating Robustness of an Acting Framework over Temporally Uncertain Domains

The ultimate goal of automated planning is the execution of plans by an artificial agent in the environment. When interactions and collaboration with humans are considered, robust plan execution requires even more highly flexible and adaptable control capabilities in artificial agents. Therefore, plan-based controllers should effectively deal with exogenous events and environment dynamics in order to perform Planning and Acting in an efficient and effective way. The general approach pursued here conforms to the general idea that Acting is not merely executing plans but it entails a more complex process in which dynamic knowledge processing and plan adaptation are required. To this aim, this paper focuses on Human-Robot Collaboration (HRC) and the timeline-based approach which is known to be well suited to robustly deal with uncontrollable dynamics. This paper presents and discusses new interesting results obtained by leveraging the acting capabilities of a novel timeline-based Planning and Acting framework called PLATINUm in a realistic HRC scenario. On the one hand, results show how the variability of the environment can negatively impact the performance and reliability of Acting systems. On the other hand, they show how a proper management of temporal uncertainty strongly improve the Actin reliability.