A Probability-Aware Planning Architecture to Support Variable Autonomy
James P. Gunderson, Worthy N. Martin · 2003
This paper proposes an architecture for autonomous systems that must be deployed in complex domains. The architec-ture is based on a probability–aware planning and execution system, which allows the system to reason about the uncer-tainties in the domain, re-evaluate plans during execution and re–plan as needed, and facilitates a non-hierarchical form of variable autonomy which allows functions from high–level, abstract planning, to low–level action selection to be passed to external controllers during plan execution. 1 Autonomous systems don’t deal well with confusion While significant strides have been made over the last 30 years in using robotics for factory and industrial automation, the use of these systems in more complex environments has