EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

4100. A novel approach for time-critical search planning in maritime search and rescue

Invited abstract in session MC-21: Disaster Response: Search and Rescue, Resource Allocation and Impactful Prepositioning, stream OR in Humanitarian Operations (HOpe).

Monday, 12:30-14:00
Room: 49 (building: 116)

Authors (first author is the speaker)

1. Yingying Gao
College of Systems Engineering, National University of Defense Technology
2. Qingqing Yang
National University of Defense Technology
3. Guopeng Song
National University of Defense Technology
4. Roel Leus
ORSTAT, KU Leuven
5. Kewei Yang
National University of Defense Technology

Abstract

In maritime search and rescue, an efficient search path is essential to maximize the success rate of the operation, typically indicated by the probability of containment and detection. This paper formulates a time-critical search problem, where a group of searchers moves through a discretized environment over time steps to locate the moving targets. Given the importance of finding drowning personnel timely, the decay factor of the survival rate is considered and incorporated into the objective function. To that end, we propose a novel approach specifically designed for this task, integrating Branch-and-Bound (B&B) with Reinforcement Learning (RL). To obtain more effective and non-myopic policies than handcrafted heuristics, we guide the branching process in path planning via RL and design a novel set representation and reward function for the bounding procedures associated with a policy, speeding up the convergence of unsupervised exploration. Our experimental results demonstrate the effectiveness of the proposed method, particularly on large realistic instances, improving upon the current state-of-the-art. The method can be generalized for different problem settings in operational decision support for maritime search and rescue.

Keywords

Status: accepted

Back to the list of papers