DECENTRALIZED SEQUENTIAL DETECTION WITH SENSORS PERFORMING SEQUENTIAL TESTS

                        Venugopal V. Veeravalli
                             Tamer Basar
                           H. Vincent Poor


                               Abstract

A decentralized sequential detection problem is considered where a set
of sensors making independent observations must decide which of the
given two hypotheses is true. Decision errors are penalized through a
common cost function, and each time step taken by the sensors as a
team is assigned a positive cost. It is shown that optimal sensor
decision functions can be found in the class of generalized sequential
probability ratio tests (GSPRT's) with monotonically convergent
thresholds.  A technique is presented for obtaining the optimal
thresholds.  The performance of the optimal policy is compared with
that of a policy which uses SPRT's at each of the sensors.