Dynamically Extending Planning Models using an Ontology
Lists
Only you see this
Abstract
In this paper we couple a deterministic planner with<br/>an ontology, in order to adapt to new discoveries<br/>during plan execution and to reason about the affordances<br/>that are available to the planner as the set<br/>of known objects is updated. This allows us to extend<br/>the planning agent’s functionality during execution.<br/>We use as an example planning for persistent<br/>autonomous behaviour in underwater vehicles.<br/>Planning in this scenario takes place in a symbolic<br/>model of the environment, simulating sequences<br/>of possible decisions. Ensuring that the simulation<br/>remains robust requires careful matching of<br/>the model to the real world, including dynamically<br/>updating the model from continuous sensing actions.<br/>We describe how our system constructs an<br/>initial state for planning, using the ontology; how<br/>the ontology is also used to determine the results<br/>of each action performed by the planner; and fi-<br/>nally demonstrate the performance of the system<br/>in a simulation, in which two AUVs are required<br/>to cooperate in an unknown environment, demonstrating<br/>that with additional reasoning the planning<br/>system is able to make new efficient choices, taking<br/>advantage of the environment in new ways.<br/>
BibTeX
@inproceedings{cashmore2015dynamically,
author = {Michael Cashmore and Maria Fox and Derek Long and Daniele Magazzeni and Bram Ridder and Valerio De Carolis and David M. Lane and Francesco Maurelli},
title = {{Dynamically Extending Planning Models using an Ontology}},
booktitle = {Research},
year = {2015},
}Loading related content...