Persim

In this project, innovative system-independent software components are to be developed, which will lead to an increase in the degree of autonomy in future unmanned or armed space missions, as well as to a higher reliability of navigation capabilities in the use of robotic systems.

The aim is to develop integrable software modules that create a virtual and highly realistic representation of the environment on the basis of data from representative sensors. Among other things, these sensors should record the mineralogical composition of materials (stones, regolith, water or ice) in the direct vicinity of the robot and thus provide a valuable supplement to the environment representation. With the help of the environment representation, potential guidelines for autonomous navigation will be passed on to the navigation software on the one hand, and on the other hand will be stored in a virtual memory for recurring situations. In this way, the experiences of long periods of use can be integrated into the system control in a sustainable way in order to learn more efficient and safer strategies for the future.

The development and demonstration of the integrable software components will be carried out on a rover with a gripper arm available at the DFKI. Coyote III, the robotic system to be used, was developed in the TransTerrA project (FKZ 50RA1301) for testing space technology in an environment representative of lunar or Martian surfaces.

The project builds on the latest innovations from DFKI's research projects and pursues the development of two key technologies: Space Resource Identification and Internal Simulations. These two technologies can be used very efficiently to create a system with unprecedented autonomous perception features and capabilities to identify and avoid dangerous situations. The development of an internal simulation will also lead to progress in the exploration of robotics reinforcement learning applications. Corresponding functions have already been identified in the literature on a theoretical level with regard to efficient use. However, the environment representation technology required for this is currently not implemented robustly enough to develop real-world applications that would benefit from such technology.
Furthermore, the project has a strong focus on reproducibility. Through a test-driven approach embedded in an agile development methodology, robust and comprehensive tests will be conducted offline (through simulation and through data set playback) and online (through field testing). The tests and the corresponding results are stored in connection with the data sets and the definition of different missions in a database, which also serves for the continuous verification of the functions.

The concept of the Digital Twin, from Industry 4.0, is a central theme that is applied to a space scenario in this project. In particular, two aspects of the Digital Twin will be developed:

1) The capabilities of robotic systems to digitise the state of the environment will be extended and
2) monitoring and improving long-term system performance based on realistic data-driven simulations and learning approaches using the collected data from sensors and software components.