ActBot: Sharing high-level robot AI scripts
C Creusot - 2016 25th IEEE International Symposium on Robot …, 2016 - ieeexplore.ieee.org
2016 25th IEEE International Symposium on Robot and Human …, 2016•ieeexplore.ieee.org
Robot behaviors are often hard-coded by skilled engineers to perform specific actions in
reaction to specific events. These codes are either high level and difficult to extend to new
domains (for example chatbot scripts that are restricted to dialogs) or low-level and difficult to
aggregate as a corpus (like robot control AIs). In this paper a general script format is
proposed for event-triggered rules definition of high-level decisions systems (dialog,
actuation decision, display, etc). Using such format, merging two robot AI can be done by …
reaction to specific events. These codes are either high level and difficult to extend to new
domains (for example chatbot scripts that are restricted to dialogs) or low-level and difficult to
aggregate as a corpus (like robot control AIs). In this paper a general script format is
proposed for event-triggered rules definition of high-level decisions systems (dialog,
actuation decision, display, etc). Using such format, merging two robot AI can be done by …
Robot behaviors are often hard-coded by skilled engineers to perform specific actions in reaction to specific events. These codes are either high level and difficult to extend to new domains (for example chatbot scripts that are restricted to dialogs) or low-level and difficult to aggregate as a corpus (like robot control AIs). In this paper a general script format is proposed for event-triggered rules definition of high-level decisions systems (dialog, actuation decision, display, etc). Using such format, merging two robot AI can be done by simply concatenating two sets of rules (like in AIML Chatbot). Our proposal format is inspired by hypergraph theory and solutions several problems found in classic decision knowledge representations. We dubbed the format ActBot (in reference to ChatBot). We discuss the required mechanisms to manage and interpret such scripts. The advantages of our system are manifold. First, the separation of the logic from the actual implementation which makes the scripts aggregable and the framework extensible. Second, an hypergraph representation with reactive rules as nodes and events as hyperedges which simplify the logic and give theoretical guaranty that any graph-based systems can be converted to the new framework. Third, the absence of naming for rules (nodes) which allows for dynamic hypergraph where the set of rules changes during execution. Our proposal simplify the writing and exchange of complex state-full robot scenarios while encouraging behaviors reused across projects.
ieeexplore.ieee.org