This uses simulated annealing to generate a transport network (a two-way distribution network with weighted population values).
Each station has a population value and the size of the crowd that wants to travel from one station to another is determined by the size of both the origin and destination stations.
Thus, a large origin station and a large destination station will have a large commuter crowd, a large origin station and a small destination station will have a medium commuter crowd and vice-versa and finally a small origin station and a small destination station will have a small commuter crowd.
The energy function is essentially a product of the total cost of the road (measured in Euclidean distance) and the overall Euclidean distance travelled by the crowd (assuming shortest path travelled).
While this is a great simplification of the dynamics of mass transit, it still captures some of the more important aspects of the sizes of commuting populations. For a more Singaporean context, this means that the model still captures how the crowd commuting from a high-density area like Serangoon to another high-density area like Dhoby Ghaut will be larger than the crowd commuting from a comparatively low-density area like Braddell to another low-density area like Woodleigh, barring of course real world interactions of rush-hour and city zoning.