150 likes | 251 Views
Tactical Event Resolution Using Software Agents, Crisp Rules, and a Genetic Algorithm. John M. D. Hill, Michael S. Miller, John Yen, and Udo W. Pooch Department of Computer Science Texas A&M University. Agenda. Tactical Event Resolution Design Architecture Genetic Component
E N D
Tactical Event Resolution Using Software Agents, Crisp Rules, and a Genetic Algorithm John M. D. Hill, Michael S. Miller, John Yen, and Udo W. Pooch Department of Computer Science Texas A&M University
Agenda • Tactical Event Resolution • Design • Architecture • Genetic Component • Rule-based Component • Results
Tactical Event Resolution • Normally a manual, ad hoc, process where the forces and combat effects on each side are tallied and the Operations officer and the Intelligence officer determine the outcome.
Problems with the Tactical Event Resolution Step • Time Constraints • Communication • Biases • Logistics • Simplification by aggregation • Ad hoc combat results
Solution • Automated support for tactical event resolution • Include biases • Track resources • Provide a configurable combat results mechanism
Design • Java-based • Event Resolution components • Genetic Algorithm • Java Expert System Shell (JESS) • an expert system shell and scripting language • supports the development of rule-based expert systems
User Actions • Create Events • Select Biases • Run analysis • Show results • Reconfigure and rerun as desired
Genetic Analysis Component • Biased Agents perform initial allocations • Maneuver bias • Massed fire support bias • Allocations are made by level and force • Force Summary • Combat Results Mechanism • Fitness monitor assigns a fitness value
Genetic Analysis (cont.) • More-fit allocations have a higher probability of being used to produce the next generation • Configurable probability of crossover • Configurable probability of mutation • Each new generation is evaluated propagated the same way • The most-fit allocation is selected
Rule-based Component • Forces allocated • Combat is resolved • Repeated until success or failure • All forces are expended unsuccessfully • Or a force mix is found that is successful • Default bias is to minimize forces used
Rule-based Details • Small number of rules needed (22) • Rules are easy to understand by a human • A point of comparison with GA approach • Can replace combat model as needed
Conclusions • Easy to setup and fast to run • Allows for “what-if” experimentation • Can playback and show intermediate steps • Gives more choices to the commander