From MOOS-DAWG'19

Talk: 02-Wilson

(:notitlegroup:)

Prev-Talk | Next-Talk | All-Talks | Talks-Sorted | MOOS-DAWG'19

Talk-02: Goal Sequence Selection for Underwater Vehicles with MOOS-IvP

Speaker: Mark A. Wilson, U.S. Naval Research Laboratory

James McMahon, and David W. Aha, U.S. Naval Research Laboratory

To be truly autonomous, intelligent agents should be able to respond robustly to unexpected situations, with minimal guidance from human operators. This capability is particularly advantageous for underwater vehicles, which may have limited or no communication with operators for long periods of time.

Effective operators may respond to unexpected situations by altering an agent's goals, e.g., when a new goal is afforded by changing circumstances, or when a previous goal becomes useless or unsatisfiable. Enabling agents to deliberate on the utility of their own goals ("goal reasoning") and to alter their goals accordingly can increase their autonomy, reliability, and trustworthiness. Underwater vehicles with goal-reasoning capabilities may be better equipped to pursue long-term autonomous missions than those without.

In this talk we describe our model for optimizing long-range sequences of goals in embodied robotic agents, as well as our integration of this model with MOOS-IvP for controlling an underwater vehicle, and associated test and application scenarios.

Categories:

Retrieved from https://oceanai.mit.edu/moos-dawg/pmwiki/pmwiki.php?n=Talk.02-Wilson
Page last modified on July 12, 2019, at 04:19 PM