Enhanced Behavioral Realism for Live Fire Targets Rick Evertsz, Andrew Lucas, Cameron Smith, Matteo Pedrotti AOS Group 580 Elizabeth Street, Melbourne, VIC 3000, Australia {rick.evertsz, andrew.lucas, cameron.smith, matteo.pedrotti}@aosgrp.com Frank E. Ritter College of Information Sciences and Technology The Pennsylvania State University, University Park, PA 16802 frank.ritter@psu.edu Rob Baker Defence Science and Technology Organisation PO Box 1500, Edinburgh, SA 5111, Australia robin.baker@defence.gov.au Paul Burns Australian Target Systems 161 Fallon Street, Albury, NSW 2640, Australia pburns@atspl.com.au Keywords: Autonomy, BDI, Cognitive Architecture, Human Behavior Modeling, Live Fire Targets, Robots ABSTRACT: Live fire training is an essential component of infantry and Special Forces training. Recent developments in target technology have created a need for more sophisticated human behavior models that can drive the targets to behave in a realistic and challenging manner. Such behavior models enable the targets to exhibit convincing tactical behavior, as well as coordinating as a team to confront the trainee with a more formidable foe. This paper describes the current status of an ongoing project to augment disparate target types with sophisticated behavioral capabilities. The underlying CoJACK™ behavior engine enables the deployment of targets that behave realistically, react in a timely fashion, and exhibit sufficient variation to ensure that the trainees cannot predict how opponents will behave. The behavior models are developed using the VBS2 environment so that they can be validated in advance of target deployment on the targets. The technological approach is presented along with two illustrative scenarios. We conclude with a discussion of the lessons learnt and the way forward, including the development of autonomous target vehicles. 1. Introduction The firing of live ammunition is a core part of infantry training. Indeed, Special Forces engage in live fire training on a daily basis to ensure that they are always at the peak of their capability. Lately, live fire training technology has expanded from simple static, pop-up or rail-based pop-up targets to indoor, projection-based systems, and more recently, fast moving, open range, mobile robotic targets. Each type of target supports particular aspects of live fire training. For example, static pop-up and rail-based targets are good for basic skills training. However, they are less effective for advanced training because target location is too predictable. With the advent of 3D, photorealistic virtual environments, projection-based systems have been developed that provide a more immersive experience in indoor, cinema ranges (Pair & Treskunov, 2006). Although providing significantly better immersion than pop-up targets, they are restricted to indoor environments, the image quality is limited (Darken & Jones, 2007), and 2D projections do not faithfully reproduce parallax effects; this means, for example, that a trainee cannot bring a concealed adversary into view by moving laterally. Over the last 20 years, the Australian Army has developed wheeled robotic targets, and the latest generation is being deployed on their live fire ranges. On flat, smooth surfaces (e.g., concrete) the targets are fast moving and maneuverable, and can be challenging to hit. However, they have limited autonomous capabilities, such as path finding, and can only exhibit simple behaviors such as scattering when one of the robots is hit. More advanced training scenarios require a much wider and more responsive behavioral repertoire than that Proceedings of the 23rd Conference on Behavior Representation in Modeling and Simulation. [online proceedings]. BRIMS Society: Centerville, OH. 1