Navigational guidance – A deep learning approach

This paper addresses the navigation problems facing many companies, including logistics couriers, and Uber, helping users find best route to multiple destinations in shortest amount of time. Navigation involving are formulated this as Directed Steiner Tree (DST) on directed graphs. We propose an end-to-end deep learning approach tackle DST a supervised non-autoregressive manner. The core our is Graph Neural Networks (GNNs) estimating whether node belongs optimal solution. Experiments conducted evaluate proposed approach, results suggest that using can effectively solve with at least 95.04% accuracy. Compared solving traditional methods, significantly improves solvability acceptable execution further explore how be applied different scenarios, such large-scale Moreover, we show smoothly several variants problem, Forest problems. In summary, shows promising implemented practice. Research limitations future directions also discussed.