State Spaces for Continuous, Spatial, and Adversarial Battle Management
Abstract
The Adversarial Continuous Time and Space Search (ACTSS) system helps a commander make better battle decisions by providing detailed, probabilistic models of future battle trajectories given possible courses of action. ACTSS models a noisy world with fluents, thereby allowing us to efficiently capture both the continuous and stochastic nature of real-world operations. Rather than relying on copious amounts of sampling to estimate future outcomes, fluents take advantage of process models that can either be solved in closed-form or can be efficiently updated recursively. We implemented a prediction system using this fluent-based representation, called ACTSS. To test ACTSS, we developed an abstract battlefield simulator, called Arena War. Arena War is a free-for-all melee, where the objective is simply to be the last one standing. The ACTSS system helps the commander to plan operations in Arena War, generating, evaluating, and monitoring possible futures. It identifies potential critical points in these futures, and it heuristically ranks the options for possible next actions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 2008
- Accession Number
- ADA481511
Entities
People
- Paul Cohen
Organizations
- University of California, Los Angeles