Situation-Aware Task Planning for Robust AUV Explorations in Extreme Environments

Achieving an accurate model of the underwater domain that captures all its complex dynamics represents a challenge. Therefore, planning solutions often involve replanning. High-level abstraction of the domain definition required for tractable planning means that the system often ignores specific properties, giving rise to states where survivability, reliability, and mission quality could be compromised. This work proposes an approach that combines planning, knowledge representation and decision making to achieve high-level mission goals that maintain mission survivability and improve robustness. We present the Situational Evaluation and Awareness (SEA) framework to bridge the highlevel planning and low-level mission execution systems. SEA maintains a dynamic evaluation of the state to provide a goal completion assessment for local recovery and global mission execution. The approach shows good performance during evaluation in different problems involving real scenarios.