Stochastic Robotic Simulation Tool

Energid Technologies proposes a game-theory inspired simulation tool for testing and validating robotic lunar and planetary missions. It applies Monte Carlo simulation within a multi-optimization environment tailored to the needs of NASA. Stochastic optimization is combined with randomized simulation to maximize multiple statistical measures of performance and calculate the parameters giving the extreme scenarios. The tool works with continuous parameters, such as mass and terrain properties, and with discrete parameters, such as lighting selection, gearing selection, and navigation parameters. It includes accurate modeling of sensors and terrain interaction using calculations performed on Graphical Processing Units (GPUs). The technique proposed is computationally expensive, but highly parallelizable, and the approach includes a design for distribution of computational burden over multiple computers, GPUs, clusters, and cloud configurations. The proposed combination of fast algorithms and game-theory-inspired statistical optimization will provide a powerful tool for NASA's use in planning missions.