Numerics of Discrete Element Simulations in Milli-g Environments: Challenges and Solutions

Abstract

JAXA scheduled a sample return mission to the martian moon Phobos, which is expected to launch in 2024. This mission features a small rover, jointly developed by DLR and CNES, which will scout the landing site of the sampling spacecraft. One of the main challenges the rover will have to face during the mission is the vastly unknown regolith surface of Phobos. Previous exploration missions, like the Mars Exploration Rover missions of NASA, showed that a rover getting stuck in loose regolith poses a huge threat to the success of planetary exploration missions. To prevent this, DEM simulations are used to optimize the rover's wheel geometry for Phobos' surface. The DLR framework for such simulations is partsival, a collision-based particle and many-body simulation tool for GPUs. Since partsival was programmed with simulations in Lunar, Martian, or Earth environments in mind, the framework has to be adapted for milli-g environments. Without any adaptations, the simulation results show severe scattering and physically unrealistic behavior. These issues can be traced back to numerical problems related to the introduction of microgravity. The low gravity requires very slow movement of the rover wheel, resulting in a very long simulation duration and therefore the necessity to compute billions of time steps. Error propagation is thus strongly favored. Furthermore, while simulations in Earth or Lunar environments can be conducted using single-precision, milli-g environments showed to require double-precision computations. This is due to the gravity influencing many (but not all) parameters of partsival's physics model, resulting in changes by orders of magnitude and loss of significance errors. partsival can be adapted to milli-g environments by adapting the macroscopic soil stiffness to allow for larger time steps while at the same time increasing the numerical precision. Thus, deterministic results can be computed while maintaining fast, efficient computation.