# Meta-Reinforced-Model-Based Planning and Fault-Tolerant Control for a Saturation Diving Decontamination Decompression Chamber

## TL;DR

This paper introduces a new control system for a saturation diving chamber to ensure safe and efficient operation under faults and high-pressure conditions.

## Contribution

A novel model-based planning and meta-learning-based fault-tolerant control framework is proposed for hyperbaric systems.

## Key findings

- The proposed approach achieves higher cumulative rewards and faster convergence in experiments.
- The system demonstrates improved robustness compared to conventional control methods.
- The framework maintains resilient performance in the presence of faults like valve offset and chamber leakage.

## Abstract

Saturation diving is the only viable method that enables divers to withstand prolonged exposure to high-pressure environments, and it is increasingly used in underwater rescue and marine resource development. This study presents the control system design for a specialized saturation diving decontamination decompression chamber. As a multi-compartment structure, the system requires precise inter-cabin pressure differentials to ensure safe decontamination and ventilation control under dynamic conditions, particularly in the presence of potential faults, such as valve offset, actuator malfunction, and chamber leakage. To overcome these challenges, we propose a novel model-based planning and fault-tolerant control framework that enables adaptive responses and maintains resilient system performance. Specifically, we introduce a trajectory-planning algorithm guided by policy networks to improve planning efficiency and robustness under system uncertainty. Additionally, a meta-learning-based fault-tolerant control strategy is proposed to address system disturbances and faults. The experimental results demonstrate that the proposed approach achieves higher cumulative rewards, faster convergence, and improved robustness compared to conventional methods. This work provides an effective and adaptive control solution for human-occupied hyperbaric systems operating in safety-critical environments requiring fail-operational performance.