Autonomous Sensors Collaboration for Moving Object Classification

Wireless Sensor Networks (WSN) perform collaborative distributed sensing of a moving object in an environment for the purpose of classification. This requires collecting measurements from as many sensors as possible to classify the object. However, there is a tradeoff between the value of information contained in a distributed set of measurements and the energy cost of acquiring these measurements, fusing them into a belief and transmitting the updated belief. To manage this tradeoff, sensor selection schemes are used. In this paper, the sensor selection scheme proceeds as a sequence of rounds as follows. The network chooses an active set of sensors at each round based on physical proximity to the object. From this set, one sensor is selected by bidding to take a measurement. This measurement is transmitted to the sensor selected at the next round. The main contribution is to dynamically optimize the classification performance for a given cost of sensing, communication and computation. The method is illustrated by an example.