Summary

Description

Energid Technologies proposes a software tool to predict robotic mission performance and support supervision of robotic missions even when environments and situations are profoundly unknown. It transcends common Monte Carlo simulations by supporting input parameters for which probability distributions are not available. Stochastic optimization is combined with randomized simulation to bound statistical measures of performance and convey the parameters giving the extreme scenarios. It also provides 3D immersive presentation of those scenarios. The act of performing multiple simulation runs in real time is enabled by the fast simulation capability provided by Energid Technologies' existing software combined with the development of new algorithms and software. The new algorithms cover path planning, scene rendering, sensor modeling, and robot-terrain interaction modeling. In the new software, automatic path planning is calculated using a combination of static and dynamic techniques. Scene rendering for sensor modeling is implemented using fast ray tracing for low-update-rate sensors and ray-tracing-validated rasterization for fast-update-rate sensors. Robot-terrain interaction is calculated through particle simulations implemented on graphics cards. For maximum performance, the new software allows distribution of randomized simulation runs over multiple networked PCs and cloud-based clusters. This combination of fast algorithms and statistical optimization offers a tool that provides new engineering insights and data. The software will be demonstrated on the example mission of searching for ice near near the southern lunar pole, giving evidence of the ability of the tool to support challenging relevant missions.