About METASAT Project
METASAT will provide a holistic and modular model-based framework to design and test software modules that target open architecture hardware, high-performance computing platforms for the space and aviation domain.
METASAT was born as a result of the need to find a solution to manage the growing complexity of new satellite designs using reliable on-board software technology. To achieve so, the consortium will develop a new model-based engineering methodology to design software modules.
To reach its vision, METASAT will leverage software virtualization layers through the use of XtratuM hypervisor on top of high-performance computing platforms based on the RISC-V open standard Instruction Set Architecture. The development of a toolchain to design software modules from high-level modeling languages for this hardware/software layer will be the project’s main focus. As a result, the consortium aims to impact in the time and cost of developing new systems, as currently costs are increasing due to the system complexity, reducing competitiveness, innovation, and potentially dependability across the industry.
Open hardware, high-performance platforms
METASAT aims at fostering the use of high-performance, open architectures in space applications, such as RISC-V based architectures of accelerators like GPUs and high-performance CPUs with acceleration features, e.g. a vector processing unit, targeting specifically the space domain. Existing on-board processors cannot provide the high-performance required for the use of expensive operations in space, such as AI. Currently the only way is to use FPGAs, but it requires manual coding which is expensive both in terms of development and verification. Other solutions, especially in the domain of open hardware, can be leveraged in order to reduce the cost and shorten the time to market. This can be achieved by reusing/adapting open hardware solutions, such as the ones based on RISC-V.
Using hypervisor-based bare metal virtualisation into the system architecture offers the capability to integrate payload software in a plug-and-play fashion, enabling more flexible and reconfigurable satellite systems. The hypervisor allows the software modules to be designed and implemented in an abstracted, virtualised environment, which isolates the software from core satellite functions and offers significant advantages for payload development efficiency, modularity, re-use, integration, and test. A virtualised software module can also be executed on different system implementations without modification. Fault tolerance can be achieved by running redundant instances of module, while fault isolation is guaranteed by the space partitioning.
METASAT will exploit the concept of Digital Twin to create a model-based toolchain that embraces all phases of the software development. In the design side, METASAT will develop and integrate a process supporting the creation
of software modules from models by automatic code generation. Open-source, freely accessible solutions for the design of software application running on hypervisors and high-performance hardware platforms will be investigated. However, the developed Digital Twin concept will consist of more than code generation tools, since it will also provide a software engineering approach that can affect the entire system lifecycle from requirements gathering to system validation. With this objective in mind, modern simulation standards will be used as building blocks for implementing a co-simulation framework.
AI-based design and testing
METASAT AI-based software testing will reduce time to market and will increase the efficiency of the organisation to produce more sophisticated software for space applications. The ability to implement fail-safe units will increase the modularisation and the sustainability by allowing the end user to compose complex interconnected systems. Overall, the use of AI for design and testing has the potential to bring the following benefits: faster identification of new alternative design solutions for future product upgrades; easy the troubleshooting by improving the time to discover faults and errors; and increase test coverage and test scope to overall improve the software quality.
On-board AI for FDIR
Although the idea of applying AI to Failure Detection, Isolation, and Recovery (FDIR) is not new, at least from a theoretical standpoint, its implementation in practice is not straightforward since AI algorithms tend to be computationally burdensome. METASAT will provide a platform able to cope with the computational requirements and a design flow to target it. On the platform side means to access hardware accelerators from the software application and the hypervisor will be implemented. On a design level METASAT will develop the tools required to model the algorithms and to deploy them on the target platform through code generation, including the programming models used by AI and general-purpose accelerators.