Abstract
Underground mining is a hazardous environment, with frequent accidents leading to significant loss of life each year. To enhance safety, sensor nodes monitor key environmental factors such as temperature, toxic gases, and miners' locations, as well as transmit critical messages. Miners interact with these sensors, which track their movements, enabling their location to be determined even without GPS signals. Therefore, predicting the battery life of these sensors is essential for: (i) rerouting miners during emergencies, (ii) ensuring timely maintenance, and most importantly (iii) identifying sensors that need energy harvesting to maintain vital communication within the mine. In this work, we propose a deep reinforcement learning (DRL) approach, Proximal Policy Optimization-Long Short-Term Memory (PPO-LSTM), specifically tailored for the mining environment. This approach considers miners' movements and communication dynamics to predict sensor battery levels, facilitating timely "energy harvesting"for sensors nearing depletion at critical locations within the mine. Our PPO-LSTM framework integrates LSTM networks with PPO to leverage temporal data correlations, enabling better decision-making for energy management. Our extensive simulations demonstrate that the PPO-LSTM framework significantly outperforms current state-of-the-art methods, including Deep Deterministic Policy Gradient (DDPG) and Soft Actor-Critic (SAC). Specifically, it achieves improvements of approximately 4%, 1.07%, and 5-10% in Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE), respectively.
Recommended Citation
M. A. Yadav et al., "Predicting Battery Levels of Sensor Nodes using Reinforcement Learning in Harsh Underground Mining Environments," Proceedings of the ACM Symposium on Applied Computing, pp. 2048 - 2057, Association for Computing Machinery, May 2025.
The definitive version is available at https://doi.org/10.1145/3672608.3707979
Department(s)
Computer Science
Second Department
Mining Engineering
Publication Status
Open Access
Keywords and Phrases
proximal policy optimization; underground mines; wireless sensors
Document Type
Article - Conference proceedings
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 2025 Association for Computing Machinery, All rights reserved.
Publication Date
14 May 2025
Comments
Centers for Disease Control and Prevention, Grant None