I am wondering if there is a body of work for merging multiple graphs into a single graph.
In my case, the graph consists of nodes which represent geo-locations (latitude, longitude) and edges which represent road connections between the nodes. I have multiple observations of different "crops" of the graph, corresponding to a list of nodes and their connections in a limited geographical area. What I am trying to achieve is to join all local graphs into a global graph. The geographical area of a sub-graph can overlap with the area of a different sub-graph.
I know that this is trivial if the sub-graphs are consistent regarding node position and edge connectivity for overlapping areas. In this case, one could simply merge sub-graphs into a single global graph by accumulating the nodes and edges.
However, I am not sure how to join such sub-graphs if the node-position and node connectivity is noisy. This means that the node positions and connections can be inconsistent when multiple sub-graphs cover overlapping areas.
I would assume that a maximum-likelihood estimation of the global graph could be somehow generated, given the fact that each sub-graph could be viewed as a noisy partial observation of the global graph. However, I have no clue where to start or whether such formulations already exist.