Open Access
Subscription Access
Cellular Maze: Cellular Network-Aware Routing System
Abstract
Cellular Maze is a machine-learning-powered navigation platform that recommends routes based on predicted cellular signal quality alongside estimated time of arrival (ETA) and distance. Conventional routing engines such as Google Maps and Waze optimize almost exclusively for travel time, which leaves users exposed to dropped calls, disconnected ride-hailing sessions, offline fleet trackers, failed emergency SOS messages, and stalled navigation updates whenever a route passes through a dead zone. The proposed system retrieves up to seven road-snapped alternatives from the TomTom Routing API, samples each path at 500 m intervals, and predicts signal strength, call-drop probability, and handoff risk at every point using ResidualSignalNet, a multi-task PyTorch model with about 560K parameters. Predictions are cross-validated with a COST-231 Hata and Ericsson 9999 propagation ensemble, combined with ETA through a user-controlled speed-versus-signal preference, and personalized over time using Thompson Sampling. Results are presented on a Mapbox GL JS map with signal heatmaps, real cell tower overlays, and predictive dead zone warnings, enabling users to stay connected without giving up control over travel time.
References
Hata, M. (1980). Empirical Formula for Propagation Loss in Land Mobile Radio Services. IEEE Transactions on Vehicular Technology, 29(3), 317–325.
European Commission, COST Action 231. (1999). Digital Mobile Radio Towards Future Generation Systems: Final Report. EUR 18957, Office for Official Publications of the European Communities.
Phillips, C., Sicker, D., & Grunwald, D. (2013). A Survey of Wireless Path Loss Prediction and Coverage Mapping Methods. IEEE Communications Surveys & Tutorials, 15(1), 255–270.
He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep Residual Learning for Image Recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 770–778.
Gal, Y., & Ghahramani, Z. (2016). Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning. Proceedings of the 33rd International Conference on Machine Learning (ICML), 1050–1059.
Thompson, W. R. (1933). On the Likelihood That One Unknown Probability Exceeds Another in View of the Evidence of Two Samples. Biometrika, 25(3–4), 285–294.
Russo, D. J., Van Roy, B., Kazerouni, A., Osband, I., & Wen, Z. (2018). A Tutorial on Thompson Sampling. Foundations and Trends in Machine Learning, 11(1), 1–96.
Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning: An Introduction (2nd ed.). MIT Press.
Luxen, D., & Vetter, C. (2011). Real-Time Routing with OpenStreetMap Data. Proceedings of the 19th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems, 513–516.
Paszke, A., et al. (2019). PyTorch: An Imperative Style, High-Performance Deep Learning Library. Advances in Neural Information Processing Systems (NeurIPS), 32, 8024–8035.
Refbacks
- There are currently no refbacks.