Darryl K. Ahner

Assignment scheduling capability for unmanned aerial vehicles: a discrete event simulation with optimization in the loop approach to solving a scheduling problem

Many military planning problems are difficult to solve using pure mathematical programming techniques. One such problem is scheduling unmanned aerial vehicles (UAVs) in military operations subject to dynamic movement and control constraints. This problem is instead formulated as a dynamic programming problem whose approximate solution is obtained via the assignment scheduling capability for UAVs (ASC-U) model using concepts from both simulation and optimization. Optimization is very effective at identifying the best decision for static problems, but is weaker in identifying the best decision in dynamic systems. Simulation is very effective in modeling and capturing dynamic effects, but is weak in optimizing from alternatives. ASC-U exploits the relative strengths of both methodologies by periodically re-optimizing UAV assignments and then having the simulation transition the states according to state dynamics. ASC-U thus exploits the strengths of simulation and optimization to construct good, timely solutions that neither optimization nor simulation could achieve alone

Optimal multi-stage allocation of weapons to targets using adaptive dynamic programming

We consider the optimal allocation of resources (weapons) to a collection of tasks (targets) with the objective of maximizing the reward for completing tasks (destroying targets). Tasks arrive in two stages, where the first stage tasks are known and the second stage task arrivals follow a random distribution. Given the distribution of these second stage task arrivals, simulation and mathematical programming are used within a dynamic programming framework to determine optimal allocation strategies. The special structure of the assignment problem is exploited to recursively update functional approximations representing future rewards using subgradient information. Through several theorems, optimality of the algorithm is proven for a two-stage Dynamic Weapon-Target Assignment Problem.

Dynamic allocation of fires and sensors (DAFS): a low-resolution simulation for rapid modeling

High-resolution combat models have become so complex that the time necessary to create and analyze a scenario has become unacceptably long. A lower resolution approach to entity-level simulation can complement such models. This paper presents dynamic allocation of fires and sensors (DAFS), a low-resolution, constructive entity-level simulation framework, that can be rapidly configured and executed. Through the use of a loosely-coupled component architecture, DAFS is extremely flexible and configurable. DAFS allows an analyst to very quickly create a simulation model that captures the first-order effects of a scenario. Although the modeling of entities is done at a low-resolution, DAFS contains some sophisticated capabilities: within the model, commander entities can formulate and solve optimization problems dynamically. DAFS can be used to explore large areas of the parameter space and identify interesting regions where high-resolution models can provide more detailed information

Data Quality Challenges in Distributed Live-Virtual-Constructive Test Environments

Live-Virtual-Constructive (LVC) simulations are complex systems comprising a combination of live (real people operating real equipment), virtual (real people operating simulated equipment or vice versa), and constructive (wholly simulated) entities. Nodes in the system support the simulation of one or more entities and are often geographically distributed to leverage unique assets (e.g., physical test range space or high-fidelity full motion simulators). Nodes are connected in a peer-to-peer fashion and communicate using protocols such as Distributed Interactive Simulation (DIS) [DIS Steering Committee 1998], the High Level Architecture (HLA) [Dahmann et al. 1997], or the Test and Training Enabling Network Architecture (TENA) [Powell and Noseworthy 2012]. Distributed LVC simulation promises a number of benefits for the test and evaluation (T&E) community, including reduced costs, access to simulations of limited availability assets, the ability to conduct large-scale multiservice test events, and recapitalization of existing simulation investments. Consequently, the Department of Defense (DoD) is increasingly turning to LVC simulation and virtual environments to support T&E events. LVC simulations have been used to test communications for unmanned aircraft systems [Parker et al. 2009], conduct cyber-security analysis [Van Leeuwen et al. 2010], and quantify radar measurement errors [Hodson et al. 2013]. Ensuring rigorous results for T&E events supported by LVC simulation requires addressing three fundamental data quality challenges: quantifying numerical errors due to weakly consistent nodes, assessing measurement accuracy with respect to tolerance requirements, and assessing measurement quality in the absence of absolute truth values.

Team task allocation and routing in risky environments under human guidance

In this paper we design coordination policies for unmanned vehicles that select and perform tasks in uncertain environments where vehicles may fail. We develop algorithms that accept different levels of human guidance, from simple allocation of priorities through the use of task values to more complex task partitioning and load balancing techniques. The goal is to maximize expected value completed under human guidance. We develop alternative algorithms based on approximate dynamic programming versions appropriate for each level of guidance, and compare the resulting performance using simulation results.

Real-Time Planning and Control of Army UAVs Under Uncertainty

With advances in sensor technology and data fusion used in military operations, more information is available for decision making. A key question is how to make efiective use of this information. One speciflc area the Army has funded is the development of Unmanned Aerial Vehicles (UAVs) as part of its Future Combat Systems (FCS). These UAVs use information from higher level sensors as cues and perform part of the reconnaissance, surveillance, and target acquisition (RSTA) mission. We develop a neuro-dynamic approach to real-time planning and control of unmanned aerial vehicles with a focus on accounting for vehicle risk and stochastic arrivals of new tasks.

Incorporating information networks into military simulations

Information superiority is considered a critical capability for future joint forces. As advances in technology continue to boost our ability to communicate in new and different ways, military forces are restructuring to incorporate these technologies. Yet we are still limited in our ability to measure the contributions made by information networks. We describe three recent studies at the Naval Postgraduate School that involve information networks. First, we examine a simulation model expanded from a two-person, zero-sum game to explore how information superiority contributes to battlefield results and how sensitive it is to information quality. Second, we examine how network-enabled communications affect the logistics operations in a centralized receiving and shipping point. The results are intended to provide operational insights for terminal node operations within a sustainment base. Third, we explore how social networks might be incorporated into agent-based models representing civilian populations in stability operations.

Using a low-resolution entity model for shaping initial conditions for high-resolution combat models

Determining the initial conditions for high-resolution combat models presents a challenging modeling problem. These initial conditions can have a major impact on the outcome of the analysis, and yet there is a significant difficulty setting those conditions in a manner that spans the important areas of the input factor space. This paper presents a method for setting those initial conditions using a low-resolution, entity-level combat model, Joint Dynamic Allocation of Fires and Sensors (JDAFS). Like its predecessor DAFS, JDAFS models entities on the battlefield, but to a lower degree of detail than most high-resolution combat models. This allows substantial exploration of the input factor space, and can help make the eventual high- resolution simulation runs more effective.

Fragment capture simulation for MANPADS test arena optimization

The assessment of aircraft survivability against explosive munitions is an expensive undertaking. Test articles for both aircraft and weapon are scarce due to their high costs, leading to a limited supply of test data. The development of newer, hopefully more effective weaponry and protection measures continues despite the short supply of test articles. Therefore, test organizations need to explore methods for increasing the quality of test results while looking for ways to decrease the associated costs. This research focuses on the Man-Portable Air-Defense System (MANPADS) as the weapon of choice and live-fire arena testing as the experimental data source. A proof-of-concept simulation infrastructure is built and used to assess and potentially optimize live-fire test arena configuration to improve missile fragment velocity estimates. Several research questions are explored: measuring potential data quality, comparing arena designs, and improving arena configurations based on fragment pattern predictions.

Weapon tradeoff analysis using dynamic programming for a dynamic weapon target assignment problem within a simulation

We consider the sequential allocation of differing weapons to a collection of adversarial targets with the goal of surviving to destroy a critical target within a combat simulation. The platform which carries the weapons proceeds through a set of sequential stages and at each stage potentially engages targets with available weapons. The decision space at each stage is affected by previous decisions and the probability of platform destruction. Simulation and dynamic programming are then used within a larger dynamic programming framework to determine allocation strategies and develop value functions for these mission sets to be used in future, larger and more complex simulations. A simple dynamic programming example of the problem is considered and used to generate a functional approximation for a more complex system. The developed methodology provides a tractable approach to addressing complex sequential allocation of resources within a risky environment.