Abstract
This paper defines and solves the sensor location problem (SLP); that is, the authors look for the minimum number and location of counting points in order to infer all traffic flows in a transport network. A few greedy heuristics are set up that find lower and upper bounds on the number of sensors for a set of randomly generated networks. It is proven that solving the SLP implies that the origin/destination (O/D) matrix estimation error always be bounded. With respect to alternative sensor location strategies, simulation experiments show that: 1) measurement costs being equal, the O/D estimation error is lower, and 2) conversely, O/D estimation error being equal, the number of sensors is smaller. (A)