Mission-driven sensor management based on expected-utility and prospect objectives

A set of mission-driven objective functions, which can be used for sensor management, is proposed. While an optimization algorithm maximizes the output of objective functions, the definition of such a function is a challenge in itself. Objectives in sensor research are usually based on sensing characteristics (e.g. estimation error, information measure), but they do not necessary describe mission goals. To improve this, we present two types of objective functions based on expected-utility and prospect. Their generality are demonstrated and their properties are compared in two case studies where the end-user has the goal to protect and/or capture physical objects using effectors that are supported by radar sensors and video cameras. A benefit is that the objectives are linked to relevant operational parameters, that are understandable for end-users, and the objectives can be fine-tuned according to end-user's preferences. Because the mission-driven objectives define on a high level what should be achieved in the end, they actually allow to utilize the employed sensors at their full potential.