Simulating Collaborative Robots in a Massive Multi-agent Game Environment (SCRIMMAGE)

Kevin J. DeMarco, Eric Squires, Michael Day, Charles Pippin

Abstract: Testing mobile robotic systems in the field is a costly and risky task. Unfortunately, there is a gap between the existing simulation capabilities and those required to simulate large numbers of aerial vehicles. Many multi-agent robotics simulators have been restricted to the 2D plane, which limits their usefulness for aerial robotic platforms. While high-fidelity 3D robotics simulators exist, simulating large numbers of agents in these simulators can result in slower-than-real-time performance. SCRIMMAGE provides a 3D robotics environment that can simulate varying levels of collision detection, sensor modeling, communications modeling, and motion modeling fidelity due to its flexible plugin interface. This allows a robotics researcher to simulate hundreds of aircraft with low-fidelity motion models or tens of aircraft with high-fidelity motion models on single computer. SCRIMMAGE provides tools for batch simulation runs, varying initial conditions, and deployment to a cluster.