Optimal routes in GIS and emergency planning application (Dunn and Newton, 1992)

This is an old paper presenting two algorithms (i.e., Dijkstra’s algorithm and out-of-kilter algorithm) for shortest path calculation. The different is that out-of-kilter algorithm can tackle the problems with flow control. This is useful in many situations such as transportation. And as the authors note, it is still far from the reality if we apply it to support real-world decisions. However, this paper seems to be limited in algorithm and its applications even it reveals the necessity to engage researchers from other disciplines (e.g. operational science). With recent development, we may need to realize that the limitation for developing better algorithms for shortest path or other optimal solutions is not about algorithms itself. It is possibly about how we construct the network, in other words, how we create an appropriate representation of the real world using arcs and nodes. It is a more fundamental level to discuss the limits of applying network analysis techniques.

Tackling questions in geography usually needs multidisciplinary knowledges, especially in the current age where the massive high-dimensional geo-tagged data (i.e., big data with both spatial and non-spatial attributes) are generated through information and communication technologies. The complexity of real-world problems dramatically increase when such massive data involve. Therefore, the challenge is we need reduce the complexity and transform these data to networks.  The design of networks should balance between efficiency (i.e., reducing the complexity) and accuracy (i.e., not losing information). Another challenge is conducting analysis on large networks that cannot be handled by old algorithms, such as traditional Dijkstra’s algorithm. The guidance for addressing these two challenges is not just from geography or some particular disciplines. I think it can be from any disciplines according to the context of actual problems. I believe that, in near future, there is no method can identify a “real” optimal solution.

Comments are closed.