Chasing the elusive sensor manager

Abstract

I n a modern aircraft, data fusion is the process by which data about the environment are gathered, coinbiaed, reasoned over, and presented to the pilot. Lkterinining which data to gather is obviously important to achieving effective data fusion. But the need for data depends OIL uncertain, interrelated and dynamic factors. This fact has pushed data-gathering determination beyond the ability of the human aitd led researchers to study structured decision-aiding systems called seiuor managers. This paper discusses sensor tnanagerneiit, focusing first O I L the problem it poses i n a modern tactical aircraft aitd then on attributes that would be desirable i n an effective sensor manager. Several techniques that offer promise are discussed.

Distributed Multisensor Fusion Algorithms for Tracking Applications

“…Computing the remaining (backward transition and marginal) probabilities follows the same procedure as for the parallel hybrid approximation, and the approximation P(k\k, ml, ml) can then also be computed according to (11).…”

“…We define rate as the inverse of the sensor sample period and resolution as the inverse of a positive definite error covariance matrix, the so called Fisher information. Most sensor management techniques have considered rate and resolution separately [7,8,11,13,15,19,20,21,22].…”

“…Sensor management systems that can balance tracking performance with system resources have been proposed to combat this problem. Such systems generate sensing actions, then prioritize and schedule those actions [11].…”

“…While a number of tracking algorithms assuming a centralized processing architecture have been developed, distributed processing architectures are more practical due to considerations such as reliability, survivability, communication bandwidth, and computational resources [4,11,12]. However, the merging of state estimates is more difficult in distributed tracking due to the loss of information inherent in forming the track estimates at the local processors.…”

“…Furthermore, each measurement imposes computational costs on a tracking system with a finite amount of processing capacity. What is needed is a sensor management technique that can balance tracking performance with available resources [11].…”