A risk-based object-oriented approach to sensor management

Abstract

Sensor systems play a critical role in providing situational awareness and threat assessment. Management of adaptive multifunction radars and sensor suites in today's combat systems is mostly based on an operator-defined set of priorities. Such a set can be very complex and requires a thorough understanding of the sensor capabilities and performance and of the operational needs. Changing such a set in strongly varying scenarios is prone to creation of sub-optimal system performance. In this paper a novel approach to assigning priorities to individual objects in the environment of a naval vessel is presented. This priority assignment is accomplished by means of dynamic evaluation of the risk imposed by each object with respect to the completion of the mission. More in particular a three-stage sensor management system is proposed that first uses the uncertainty related to each object's attributes, such as the state vector, classification and identification to determine the allocation of surveillance, track or recognition tasks. The sensor manager then selects the most appropriate sensor for the task and finally distributes the available time budget based on priorities in case of multiple tasks. The risk is estimated by a dynamic Bayesian network modeling relevant operational knowledge. The nodes in this DBN represent the different object states in time and the events leading to them. A computer simulation was developed to prove the viability of this concept.