The design and evaluation of mission-critical functions such as Entry, Descent and Landing (EDL) for the Mars Science Laboratory (MSL) requires large-scale numerical Monte-Carlo simulation incorporating hundreds of parameters and nonlinear models. Such physics-based models of the environment, spacecraft and on-board algorithms were successfully exercised for the Mars Exploration Rover (MER) mission even though it proved to be a major problem managing the trajectory and event data outputs from thousands of simulation runs, configuration control of the many input cases and their variants, and retrieval of specific information from the data generated. Even more complex models with a much larger set of parameters is required for MSL because of the added complexity of closed-loop guided flight, precision navigation and controlled soft landing of the rover payload directly onto terrain. Fundamentally new ways of simulation data management are therefore required for MSL.
To address this need to generate and manage large sets of data in a configuration-controlled manner for design space exploration, system performance assessment, and design optimization, MSL has implemented (using Mars focused technology funds), a simulation database manager that incorporates a relational database (using mySQL). User configurable output trajectory and event capture specifications, utilities to retrieve data from the database for simulation setup and deposit of the results back into the database, and data query tools (using PHP scripts) have been demonstrated on workstations with a single cpu/file-system hosting the database implementation, and batch jobs concurrently executing on multiple workstations.
We are adapting the MSL Simulation Data Manager toolkit for execution on the JPL supercomputer and demonstrate its operation on simulations using the Dsends (Dynamics Simulator for Entry, Descent and Surface landing) EDL simulation tool. Dsends is a simulator from the Darts/Dshell family of physics-based simulators and is one of the two primary engineering simulation tools for use on MSL EDL. Dsends shares the same Python-based middleware as the Roams rover simulator that has already been ported to the JPL supercomputer.