Multi-Agent Simulation Approach to Development of Applications for Decentralized Tactical Missions

Komenda, Cap, Pechoucek

agents simulation tactical coordination

@inproceedings{komenda:ksco-2012,
  author={Anton\'{i}n Komenda and Michal \v{C}\'{a}p and Michal P\v{e}chou\v{c}ek},
  title={Multi-Agent Simulation Approach to Development of Applications
         for Decentralized Tactical Missions},
  booktitle={Knowledge Systems for Coalition Operations},
  year={2012},
}

[ Download PDF ]

High fidelity simulation of robotic assets enables efficient experimentation, evaluation of AI controllers alongside some amount of realism

Platform enables decreasing level of abstraction

• Progressively test software in more and more detailed problems/environment
• E.g., moving from discrete motion over graph to continuous motion in virtual world

Example domain tasks:

• Patrolling allied convoys
• Capturing evading adversaries
• Low-level formation maintenance
• Coordinated observation

Conservative adaptation preserves mathematical properties (soundness, completeness, etc), but addresses domain specific constraints

Multiple kinds of state storage

• Could be organized by entity/component type
• Or organized by data type, e.g., graph, key/value, btree
  • Optimized techniques for caching, distribution, etc

Event based time management