Decision-Time Learning and Planning Integrated Control for the Mild Hyperbaric Chamber
Plateau hypoxia represents a type of hypobaric hypoxia caused by reduced atmospheric pressure at high altitudes. Pressurization therapy is one of the most effective methods for alleviating acute high-altitude sickness. This study focuses on the development of an advanced control system for a vehicle...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-06-01
|
| Series: | Algorithms |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1999-4893/18/7/380 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Plateau hypoxia represents a type of hypobaric hypoxia caused by reduced atmospheric pressure at high altitudes. Pressurization therapy is one of the most effective methods for alleviating acute high-altitude sickness. This study focuses on the development of an advanced control system for a vehicle-mounted mild hyperbaric chamber (MHBC) designed for the prevention and treatment of plateau hypoxia. Conventional control methods struggle to cope with the high complexity and inherent uncertainties associated with MHBC control tasks, thereby motivating the exploration of sequential decision-making approaches such as reinforcement learning. Nevertheless, the application of sequential decision-making in MHBC control encounters several challenges, including data inefficiency and non-stationary dynamics. The system’s low tolerance for trial-and-error may lead to component damage or unsafe operating conditions, and anomalies such as valve failure can emerge during long-term operation, compromising system stability. To address these challenges, this study proposes a decision-time learning and planning integrated framework for MHBC control. Specifically, an innovative latent model embedding decision-time learning is designed for system identification, separately managing system uncertainties to fine-tune the model output. Furthermore, a decision-time planning algorithm is developed and the planning process is further guided by incorporating a value network and an enhanced online policy. Experimental results demonstrate that the proposed decision-time learning and planning integrated approaches achieve notable performance in MHBC control. |
|---|---|
| ISSN: | 1999-4893 |