果冻传媒

ABSTRACT

RIPPIL investigates the applicability of interactive learning to enhance?automated route planning. We developed a capability that learns 果冻传媒?exploits a user’s latent reward function through pairwise comparison?feedback on presented aerial routes created using a mature, fielded?automated route planner. Initial results with synthetic reward functions?show that the presented routes improve over time (i.e., more closely align?with the feedback) 果冻传媒 the generation of route pairs using active strategies?speeds up the interactive learning process.

?

WHY?

Many automated mission planning tools have been developed 果冻传媒 matured over the past couple of decades, but often do not capture?everything a user desires, leading to non-use or time-consuming manual alteration of plans. By combining the computational strengths?of these automated planners with the expert user’s intuition 果冻传媒 domain expertise, we strive to achieve higher quality plans in a shorter time.

?

WHAT?

RIPPIL applies preference learning to automated low-observable aerial route planning by?iteratively modifying route planner inputs 果冻传媒 collecting user comparison feedback on presented?route pairs. While pairwise preference learning has been implemented on low-dimensional?problems in the literature, it has not been extensively evaluated for high-dimensional input?spaces 果冻传媒 on real user preference data. We will collect 果冻传媒 analyze data from user experiments?to determine the benefits of interactive learning for aerial route planning.

?

HOW?

RIPPIL’s Gaussian Process-based?preference learning prototype has?been tested using synthetic reward?functions to simulate preference?decisions, 果冻传媒 will be tested with?real user preferences.

?