Bitte wählen Sie ihr Lieferland und ihre Kundengruppe
In order for simulation based training to help prepare warfighters for modern asymmetric tactics, opponent models of behavior must become more dynamic and challenge trainees with adaptive threats consistent with those increasingly encountered by the military. In this paper we describe an adaptive behavior modeling framework designed to represent asymmetric adversaries within a multi-player virtual environment. The framework aims to provide a means for adversary models to analyze the tactical situation during execution, and adapt their behaviors and tactics accordingly. Dynamic adaptations occur both within an exercise and across exercise runs, with an automated means to carry 'lessons learned' forward from one exercise to the next and adapt tactics in subsequent training sessions. This paper provides details on two distinct areas of investigation. The first area is a survey of the space of asymmetric tactics and adaptations from real-world military operations, initially focusing on urban 'presence patrols'. A number of training experiments were conducted in a virtual environment to solidify the behavior modeling requirements for this specific operational area, and provide a basis for generalizing to other domains. The second research area is the design and development of artificial intelligence techniques for creating adaptive adversaries. The approach makes use of an authoring tool for defining adaptive behavior models specified as partial plans that can be instantiated with choices partly driven by reward functions using data from previous events. Based on this initial behavior specification, new adaptive behaviors can be automatically generated with methods based on evolutionary algorithms. In both cases, the adversary model adapts over time in conjunction with training events.