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

Talk-07: Platform Autonomy Meets Mission Autonomy

Zach Duguid and Mike Benjamin, Computer Science and Artificial Intelligence Laboratory, MIT

To achieve highly robust and adaptive behavior with unmanned marine vehicles, multiple levels of autonomy are required. Platform autonomy is necessary for lower level planning such as trajectory following in the presence of environmental disturbances and multiple vehicle coordination such as collision avoidance. On the other hand, mission autonomy is necessary for higher level planning such as designing a route sequence that maximizes data collection while respecting rendezvous deadlines and designing contingency plans such that the activity schedule can be adjusted during execution. To showcase the value of a robotic platform that combines platform autonomy and mission autonomy, we describe a framework for integrating model-based planning paradigms such as the Planning Domain Definition Language (PDDL) with the MOOS-IvP system and we test our approach on a simple deploy- ment scenario on the Charles River. Specifically, we explore the selective Traveling Salesman Problem (TSP) with rendezvous constraints. At the mission start, the vehicle receives a set of science targets along with rendezvous points with corresponding time windows for rendezvous. To complete this mission, a high level plan that maxi- mizes the number of science targets visited while respecting the rendezvous constraints must be devised. The high level plan should include contingency plans for when unanticipated disturbances such as adverse water currents or human-commanded station keeping prevent timely execution of the activity sequence. Once the high level plan is designed, a lower level planner must design and execute a dynamically feasible control sequence while performing state estimation. During execution, it may be deemed necessary to pursue one or more of the contingency plans depending on the state of the vehicle and the constraints of the mission. While relatively simple to describe, this deployment scenario is rich in terms of the autonomy capabilities exercised. To execute this mission successfully, it is necessary to combine platform autonomy with mission autonomy.

Categories:

  • USVs
  • Mission Planning
  • Interoperability