Prev-Talk | Next-Talk | All-Talks | Talks-Sorted | MOOS-DAWG'24

Talk-11: MOOS-IvP + Adaptive Control and Multi-Agent Decision-Making

Tyler Paine, MIT Mechanical Engineering

This talk will cover two topics:  First we will present our autonomy stack used for the DARPA project named Learning INtrospective Control (LINC) where behavior-based autonomy, MOOS-IvP, is combined with advanced vehicle controllers that learn changes to vehicle dynamics in real time and adjust the control to maintain good performance.   Secondly, we will present a new framework for multi-agent decision-making that extends MOOS-IvP by enabling the wider group to influence an individual agent's decision to select a MODE in the helm (and thus activate behaviors).  This new feature allows the group to make collective decisions in response to changing situations during a mission.   We will discuss a few experimental results that demonstrate both of these topics, and also discuss the demo we will present on Demo day.


Categories:

  • Multi-Agent Decision-Making
  • Adaptive Control