In order to ensure service continuity of connected cars moving inside a Fog Computing infrastructure under a Service Level Agreement, a service needs to migrate from a fog node to another. An approach is to always keep migrating the service towards the fog node that is the closest to the current position. However, frequent service migrations have a migration and network cost. Intuitively, the more migrations are triggered, the bigger this cost is. In this work we look into ways to reduce this cost by studying how to minimize the number of VM migrations triggered. We introduce a general case in which we minimize a linear combination of the infrastructure cost and the number of service migrations given statistics on the routes taken by the vehicles. This problem can be represented as a bipartite graph where the minimization problem is an instance of the Weighted Set Cover problem. For a special case of pair-wise mobility model in which the origin and destination of vehicles are in the coverage range of adjacent base stations, we first present a static offline ILP formulation of the migration minimization problem. For this simple case, we then propose two heuristics inspired by the greedy algorithm for the weighted set cover problem as polynomial approximations.