Case-Based Behavior Recognition in Beyond Visual Range Air Combat.

摘要

An unmanned air vehicle (UAV) can operate as a capable team member in mixed human-robot teams if it is controlled by an agent that can intelligently plan. However, planning effectively in a beyond-visual-range air combat scenario requires understanding the behaviors of hostile agents, which is challenging in partially observable environments such as the one we study. In particular, unobserved hostile behaviors in our domain may alter the world state. To effectively counter hostile behaviors, they need to be recognized and predicted. We present a Case-Based Behavior Recognition (CBBR) algorithm that annotates an agent s behaviors using a discrete feature set derived from a continuous spatio-temporal world state. These behaviors are then given as input to an air combat simulation, along with the UAV s plan, to predict hostile actions and estimate the effectiveness of the given plan. We describe an implementation and evaluation of our CBBR algorithm in the context of a goal reasoning agent designed to control a UAV and report an empirical study that shows CBBR outperforms a baseline algorithm. Our study also indicates that using features which model an agent s prior behaviors can increase behavior recognition accuracy.