Reinforcement learning in risk management for pharmaceutical construction projects: frontiers, challenges, and improvement strategies

The intelligent construction of pharmaceutical facilities faces dynamic and nonlinear risks, and traditional management methods struggle to meet the high demands for real-time response and compliance. However, the existing reinforcement learning (RL) research in this field still lacks systematic application architecture and industry governance considerations. Therefore, this paper reviews the practical applications of six algorithms—Deep Q-Network (DQN), Deep Deterministic Policy Gradient (DDPG), and Proximity Policy Optimization (PPO)—in construction safety, temperature control, resource scheduling, and automated equipment optimization, validating the potential of reinforcement learning to effectively manage dynamic risks through adaptive learning. Simultaneously, this paper accurately identifies key bottlenecks in current applications: the fidelity gap between the simulation environment and actual medical regulations, the lack of standardized reinforcement learning deployment procedures, and the ambiguity between algorithmic decision-making authority and human oversight responsibility. To address these issues, this paper pioneers a high-fidelity environment simulation scheme integrating multiple technologies, a standardized reinforcement learning application framework, and a clear rights and responsibilities governance system, providing crucial theoretical support and practical pathways for constructing a reliable and efficient paradigm for pharmaceutical facility construction risk management.

Text 124575.pdf - Published Version Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (5MB) Official URL or Download Paper: https://www.sciencedirect.com/science/article/pii/...

Actions (login required)

View Item