Space Cloud: From a Distributed On-board Computer to a Federated System-of–Systems in Space

Kurzfassung

Future space missions face enormous challenges in the area of on-board data processing; especially in the area of earth observation and robotics. Powerful and smart on-board processing becomes more and more important to cope with limited communication bandwidth to the ground and to enable higher degrees of autonomy for deep space robotic missions. The German Aerospace Center (DLR) is conducting the research activity OBC-NG (On-board Computer – Next Generation) to build a new distributed on-board computer architecture that allows sharing of computing resources across subsystems of a spacecraft. Redundancy is not limited to one particular subsystem in this architecture. Tasks of failing computing nodes can be migrated to any available computer on the spacecraft. Additionally, it is possible to reassign computing resources to specific tasks if the mission profile changes. For instance, if a probe is going to land on a celestial body by utilizing an optical navigation system, the majority of the on-board computing resources are used to run the complex computer vision algorithms during descent and landing and calculate a navigation solution and control the spacecraft. Afterwards, the computing resources are reassigned to other tasks, e.g. for scientific experiments on the ground. Both scenarios, task migration in the case of failures and changes for different mission phases, are made possible by employing the concept of reconfiguration of hardware, software, and network routing tables. The OBC-NG architecture consists of reconfigurable nodes connected via a SpaceWire network. The nodes can be realized on different computer architectures, for instance CPUs and FPGAs. A special OBC-NG middleware provides means for error detection and reconfiguration services, like checkpoint, task migration, and task morphing (i.e. moving a task, for instance, from an FPGA to a CPU). The middleware supports diverse operating systems. Real-time operating systems like RODOS running hard real-time tasks for the avionic subsystems or soft real-time operating systems, like Linux, that provide many third-party libraries for complex data processing algorithms. Dedicated interface nodes distribute sensor data to the network and provide services to access actuators of the spacecraft from any computing node in network. The next step will extend this architecture to swarms of spacecraft or robotic exploration devices like rovers. We call this concept “Space Cloud”. For this, the network is then not only limited to a single system but extended to a system of systems. In such a setup, the tasks can be distributed across several systems by running the same middleware and providing high-speed communication links. One possible scenario would be that after the landing of the aforementioned probe, several small rovers and robotic crawlers are deployed with limited computing power to explore the area and the main probe provides computing resources via a communication link for these vehicles. Another possible scenario would be a swarm of satellites which act as a distributed network of computing nodes. Tasks can be shared across several spacecraft. In summary, this contribution presents the architecture of OBC-NG and gives an outlook towards the vision of an OBC-NG-based Space Cloud architecture.