System-level diagnosis for WSN: A heuristic

Wireless Sensor Networks are distinguished by two main characteristics: they are both prone to failures, mainly those encharged of ambiental monitoring, and aware of energy consumption. We propose to address these challenges by means of a fault tolerant approach, known as system-level diagnosis. In this approach, a subset of the sensors in a network is asked to perform mutual tests among themselves in order to achieve diagnosis, more specifically, to know which t sensors in a set of n, where n ≥ 2t + 1 are in a faulty state. The set of tests to be performed by the sensors involved in diagnosis is a connection assignment. An optimal connection assignment has n = 2t +1 nodes. In previous works, it has been proven that the Optimal Design Testing Assignment, ODTA approach, is optimal under the assumptions of diagnosability of a system. Nevertheless, in order to achieve optimality in terms of minimum energy consumption, the sensors that take part into diagnosis must be choosen from the vicinity of the sensors under suspicion, which is a problem supposed to be computationally intractable. We present an heuristic to address this challenge and compare it with the optimal solution obtained through linear programming.