Tenders are invited for Advanced Heterogeneous Inference Data Processing Module - Expro plus The objective of this activity is to develop a modular data processing module for missions that will require extreme processing power, high speed serial interfaces and AI acceleration, to produce on board data products or to handle very high data rates.Description: In the space embedded systems domain, we are now witnessing a very novel trend. While "classical processing ASICs keep suffering of the usual, widening, performance gap between "terrestrial" processors and space grade ones, on the counter recent developments especially in SRAM-based FPGAs have brought "edge" technologies for space use, providing a possible quantum leap for processing in space.One of the most notable novelties is the availability of some Commercial Off the Shelf System-on-Chip components also in space-grade packages, which deliver previously impossible application and system-level performance for space edge computing applications. Being based on 7nm node, they are a huge step from previous 65 and 45 nm space grade "pure FPGAs since they also include heterogeneous compute engines with a breadth of hardened memory and interfacing technologies for far superior performance/watt targets.Especially for payload data processing tasks, these devices will likely replace any future dedicated ASSP and call for availability of sophisticated, RHBD, processor IP cores for specific applications (like Machine Learning acceleration) and for dedicated HW/SW codesign, use of advanced OS (like Linux). The engineering challenges in using these components in space are very demanding: although radiation data show that for example some 7 nm technology will provide SEL and SET-free devices, management of soft errors, design of stable GHz clocks and high speed links, design of a power distribution system capable of delivering e.g. 100 A @ 0.8 V, PCB design and thermal dissipation design are all very critical.The activity will identify a target use case and perform all the design activities starting from architectural design identifying the main building blocks. A survey on related implementation metrics shall also be performed (e.g. from the system requirements, user needs, the choice of data handling architecture then FPGA package, IP cores gate count, performance and interface requirements, DDR and NVRAM memory, storage capacity, high speed serial links, max dissipated power, form factor etc.). This shall consider not only the functions that will be supported in the current activity but also foreseen extensions (e.g. DNN acceleration implementation in programmable logic, direct RF payload processing), which will restrict the list of available selections.The final development of a flexible, reusable and cost-effective generic processing module for institutional. And commercial satellites (150+ kg class) in Low Earth Orbit (LEO) is aimed at accelerating machine learning processing functions. The software infrastructure will be designed as an open environment potentially by passing usual space software quality checks, to facilitate the evaluation and testing of newly developed on-board applications through end-to-end system scenarios. The software suite is expected to be based on opensource applications, such as Linux OS, OpenCL, OpenMP, PyTorch, etc. The goal is to create a building block suitable for the Satellite-as-a-Service (SaaS) concept. Moreover, the activity will also foresee the definition of a development roadmap for deploying the technology on planned and future IOD mission(s) for concerned validation campaigns in the frame of an E2E system scenario.The following tasks are foreseen within this activity: Identify specific use cases and derive detailed requirements. Analyze requirements and trade-offs to determine the optimal architectural design and functional allocation. Develop detailed designs for the electronic module in EQM quality and associated ground support equipment (EGSE). Implement the design, including the selection and integration of FPGA, processor IP cores, memory, and other components. Manufacture, assemble, and integrate the EQM module into a relevant unit environment. Conduct thorough testing of the module and its EGSE. Develop a roadmap for deploying the technology on future In Orbit Demonstration (IOD) missions, enabling validation campaigns in end-to-end system scenarios.Deliverables: Reports, EM module and relevant EGSE.Read more Tender Link : https://esastar-publication-ext.sso.esa.int/ESATenderActions/filter/open
