Information Based Sensor Management (IBSM) was briefly described in an earlier posting, but the question arises as to what are the criteria which led to this inevitable approach to sensor management. That is, how do you architect the noncommensurate elements of IBSM while satisfying competing mission goals. We present here a brief synopsis of an invited paper by the same title which will be presented at the March 22-24, 2021 Association for the Advancement of Artificial Intelligence (AAAI) Spring Symposium, SSS-21.
The sensor management system design goal is to provide the decision maker with a maximum mission valued, minimum uncertainty model of the world in a timely manner through the effective use of heterogeneous, multi-modal sensors operating in physical, social, and cyber space on one or more platforms. We do this by designing a system of systems (SOS) which maximizes the transfer of information from the real world into our model of the world. No longer is it sufficient to design data acquisition systems which operate at their maximum data acquisition capability, but rather selectively acquire that data which provides mission valued information. Therefore it is important to recognize that intelligence, surveillance, and reconnaissance (ISR) sensors are operating under constraints in measurement, computation, or data space. So what are some of the criteria which can be applied to maximize the usage of the available sensors?
Viewing sensors as communications channels guides one into maximizing the expected information value rate (EIVR) from the real world, through the sensors, to a probabilistic world model along with the real uncertainties. Given this underlying criterion, other constraints need to be incorporated such as putting the human on the loop (HOL) rather than in the loop (HIL) because of the limited bandwidth of a human operator when applied to directly control sensors. It is further desirable to build the SoS of orthogonal components such that their interaction is predicated on a few well-defined interfaces rather than an ad hoc interconnect structure. If this is the case, then individual subsystems can be improved or replaced without a complete redesign of the system. Furthermore, one must carefully differentiate between the data acquisition component and the detailed sensor scheduling, the information extraction, the representation of the world as a probabilistic model, and the translation of information needs into real, measurable sensing functions.
While many attempts have been made to combine noncommensurate performance measures into a single value, generally through a weighted sum of non-dimensional components, it is far more effective to use a mission valued goal lattice to weight individual situation information needs and individual sensor information alternatives. The goal lattice approach also enables the HOL control through the modification of relative goal values.
Click here for more information or to order.
