MinerFinder: A GAE-LSTM method for predicting location of miners in underground mines Abhay Goyal Dept. of CS, Missouri S & T, USA aghnw@umsystem.edu Sanjay Madria Dept. of CS, Missouri S & T, USA madrias@mst.edu Samuel Frimpong Dept. of CS, Missouri S & T, USA frimpong@umsystem.edu ABSTRACT Recent reports by the Mine Safety and Health Administration sug- gest that several injuries and fatalities could be attributed to the inability to accurately locate miners in case of disasters. Since under- ground mines have a complicated geometrical landscape and tech- nological constraints such as no GPS information available, it is dif- fcult to predict the location of a miner and hence may cause delays and inefciencies in rescue operations during a disaster. A signif- cant amount of research has been done to capture complex spatio- temporal relationships of movement of the nodes/people/things with time, spatial and temporal features to separately extract these relationships for location prediction. Although Markov Chains (MC) and Recurrent Neural Network (RNN) based methods have been used to predict locations, not all of them specifcally mention the spatial locations, their connections and the aggregation techniques which would allow for the actual representations of the trajectory of miners. Addressing these concerns, we develop a frst-of-its- kind end-to-end system entitled MinerFinder to predict the future location of the miners by incorporating Long Short Term Mem- ory (LSTM) for trajectory information with Graph Autoencoder (GAE) for spatial environmental information representing the node connectivity. In addition, our approach will combine the miners’ previous trajectories and daily repetitive patterns enhancing the prediction robustness. We evaluated MinerFinder over synthetic dataset to analyze the structure and location topology of an un- derground mine compared with foreground locations. Our model outperforms state of the art models and achieves an AP score rang- ing from (0.62 - 0.68) and Receiver Operating Characteristics (ROC) ranging from (0.63-0.68) with increasing percentage of prominent locations (most visited) to 50%. CCS CONCEPTS · Computer systems organization → Embedded systems; Re- dundancy; Robotics; · Networks → Network reliability. KEYWORDS Graph Autoencoder, LSTM, Location Prediction, underground mine ACM Reference Format: Abhay Goyal, Sanjay Madria, and Samuel Frimpong. 2022. MinerFinder :A GAE-LSTM method for predicting location of miners in underground mines. Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). SIGSPATIAL ’22, November 1ś4, 2022, Seattle, USA © 2022 Copyright held by the owner/author(s). ACM ISBN 978-1-4503-9529-8/22/11. https://doi.org/https://doi.org/10.1145/3557915.3561024 In The 30th International Conference on Advances in Geographic Information Systems (SIGSPATIAL ’22), November 1ś4, 2022, Seattle, USA. ACM, New York, NY, USA, 12 pages. https://doi.org/https://doi.org/10.1145/3557915.3561024 1 INTRODUCTION Disaster management in underground mines can be better handled using improved information and communication management ser- vices to help locate trapped miners during a disaster 1 . The leaky feeder system [[1], [2], [3]] is the most prominent method used for locating miners, and has been used for a long time. But in case of a disaster, the whole system of wired communication and the location prediction both are disrupted [[4][5], [6]]. Note that the localization using Zigbee, Wi-Fi, and RF-based trilateration and triangulation cannot be accomplished in underground mine due to the infrastructure unavailability during a disaster when it is needed the most. Even then, most of the indoor localization technologies in underground mine are not appropriate and perform poorly [7] due to several technological constraints and dense device deploy- ment requirement not possible in the underground mine (due to the complex landscape). One possible solution in such scenarios is the use of Delay Tolerant Networks (DTNs) communication system to locate miners as DTN does not need any communication infras- tructure inside the mine. A DTN system has the ability to collect and disseminate real-time information using wireless communica- tion about the miner’s location and the environment when nodes connect. Thus, during a disaster, using the data collected by DTN devices, the system can predict every miner’s locations, which can save them in case an evacuation is needed. Below we enumerate the additional challenges faced in deter- mining the location of miners in an underground mine: • Since there is no GPS information available in underground mines, miners use the method of navigation based on pillars (red color boxes in Fig. 1). Thus, location prediction in un- derground mine refers to predicting the locations (pillars) of miners. • The high attenuation posed by the pillars are a great chal- lenge for communication in a mine. These pillars, often high and wide, do not allow for communication signals to pass through. Due to the attenuation of signal underground, line- of-sight is the only method to communicate, and hence, not much data can be communicated through DTN interactions. • Lack of a reliable wireless tracking systems to locate miners due to unreliable power cables, and high powered battery device is not allowed inside due to fammable gases inside. Fig.1 shows how the DTN nodes are connected to each other and use their data <miner/DTN ID, pillar number, time, speed, angle> to predict locations (with respect to pillars) of nodes. Some 1 https://wwwn.cdc.gov/NIOSH-Mining/MMWC/MineDisasters/AccidentType