SBIR-STTR Award

The safety and operational success of the U.S. Navy (USN) depends on expert navigation, seamanship, and shiphandling skills. Tragically, the Navy experienced four major incidents in 2017. The resulting USN Comprehensive Review identified lapses in basic seamanship and safe navigation skills as contributing factors, reinforcing the critical need for rigorous shiphandling training and proficiency assessment. Virtual Environment (VE) training provides a promising and cost-effective way to build such proficiency. While demand is high, VE training heavily relies on instructor coaching from master mariners. By developing more advanced intelligent tutoring systems (ITS) that augment this process, instructors can manage more student trainees and students can be exposed to a wider variety of task scenarios with tailored feedback. To address this, we propose to design and demonstrate a system for Mentoring and Responsive Learning through Intelligent Nautical Skill-modeling, Prompting, Intervention, and Feedback during Instructor-Controlled Exercises (MARLINSPIKE). MARLINSPIKE provides a skill-centered ITS framework for managing training based on real-time skill assessment through a combination of improved performance tracking, automated feedback, alerting methods, and automated assessment data capture for debriefings. The resulting system will increase the effectiveness and efficiency of VE-based training methods while reducing demands for instructors.

Benefit:
We expect the full-scope MARLINSPIKE program to have immediate and tangible benefits for Surface Warfare Officer student trainees and their instructors across the Navy. MARLINSPIKE will provide robust intelligent tutoring and other novel support tools to instructors in virtual training environments. We see a commercially viable market licensing or selling MARLINSPIKE to enhance simulation-based training systems across military and non-military environments. MARLINSPIKE will also contribute several technical improvements to our MAST software, increasing its appeal as a commercial product.

Keywords:
Instruction Management, Instruction Management, Virtual Coach, Training performance assessment, Skill Trees, Open System Architecture, Director Agents, Intelligent Tutoring System, Conning Officer Virtual Environment (COVE)

The safety and operational success of the US Navys (USN) fleet depends on the expert navigation, seamanship, and ship handling skills of its Surface Warfare Officers (SWOs). To address these challenges, Charles River Analytics and University of Nevada Reno (UNR) propose to develop and evaluate a full-scope prototype Mentoring and Responsive Learning through Intelligent Nautical Skill-modeling, Prompting, Intervention, and Feedback during Independent Exercises (MARLINSPIKE) system. MARLINSPIKE features (1) a scenario generation and director agent that uses search-based experience management to project potential training outcomes and plan training scenario elements from high-level, quick-to-author specifications; (2) Charles River Analytics Hap reactive planning architecture to make realistic and reactive high-level decisions for each ship; (3) UNRs ship movement controller using rules of the road and potential fields to navigate ships through traffic and around obstacles to meet the high-level decisions; and (4) UNRs Rules of the Road ship handling simulation to enable ease of integration and rapid iteration on development prototypes, transitioning to Navy MSS and NSST simulations for transition efforts.

Benefit:
We expect the full-scope MARLINSPIKE system to immediately and tangibly benefit target users across the DoD. In particular, MARLINSPIKE will provide intelligent ship driving behavior models and training scenario management. Incorporating the innovations developed under MARLINSPIKE will reduce time and costs of ship handling virtual training while increasing availability, variety, and standardization. In the private sector, we see a commercially viable market licensing or selling MARLINSPIKE as a solution for private ship handling. We will also incorporate advances in Hap made under this effort to enhance our Believable Agent Custom Solutions software, increasing its appeal as a commercial product.

Keywords:
simulation-based training, Reactive planning, Virtual Environments, Potential Fields, Training, Ship handling, Behavior Modeling