The global marine industry is transitioning to smart ships where navigation, maintenance, and operations are done autonomously. Integrating autonomy into already complex ships presents many challenges, including identifying faults and taking corrective actions. These actions are key components in a Self-Adaptive Health Monitoring (SAHM) system which aims to maintain ship operations. The challenge lies in the failure state space’s extraordinary size which current methods aren’t capable of dealing with. Diagnosis has been achieved for smaller scale systems such as NASA deep space probes, but the complexity of a probe is equivalent to a ship’s single small sub-system. The authors combine recent advancements in statistical physics and multi-agent-based reinforcement learning to address the scale issue and enable crewless vessels. Statistical physics works to extract information about objects through tensor networks, combining physical and logical representations of objects. By combining a sequence of contractions, an ensemble of data about the physical system can be constructed quickly. To demonstrate the proposed method, the algorithm is applied to a modified version of the N-Queens problem which contains operational decision making, geometrical constraints, and a scalable problem. The authors then apply an already proven method to the modified version of the N-Queens problem and compare the results. The tensor network enables agents to handle state space explosion by decoupling the system’s complexity from decision making.
Diagnosis Fault detection Self-adaptive health monitoring Tensor networks
Office of Naval Research
N00014-17-1-2491
Thanks to Ms. Kelly Cooper of the Office of Naval Research
N00014-17-1-2491
Birincil Dil | İngilizce |
---|---|
Konular | Yapay Zeka, Bilgisayar Yazılımı |
Bölüm | Research Articles |
Yazarlar | |
Proje Numarası | N00014-17-1-2491 |
Erken Görünüm Tarihi | 30 Temmuz 2023 |
Yayımlanma Tarihi | 1 Aralık 2023 |
Kabul Tarihi | 3 Temmuz 2023 |
Yayımlandığı Sayı | Yıl 2023 Cilt: 3 Sayı: 2 |
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.