Performing on-board data processing in the spacecrafts typically requires large amount of computing capabilities to handle the large data rates that the edge computing applications require. These applications can be for instance image processing for Earth observation, automated landing, and on-board decision making for debris avoidance. More and more of these applications are relying on Artificial Intelligence rather than traditional computing to analyze the data and make decisions. Because of the high processing requirements and the specificities of executing AI, it is not only a question of using the fastest processors, but also of optimizing the resource utilization when computing the neural networks. Another requirement that comes on top of the pure computing performance is the capability to operate multiple tasks (i.e. neural networks) in parallel to manage several functions and possibly serve several users. This is served both by multicore processors to split the tasks among different cores, and by implementing the software in a smart way to share efficiently the resources (CPU usage, memory, …).
Klepsydra is a Swiss SME focusing on developing high-performance edge data processing software for autonomous embedded systems to serve applications in the Space, aerospace, smart mobility and IoT industries. Klepsydra will provide the high-performance SW framework that allows to perform the necessary computing and AI/ML tasks on the available on-board computer.
Teledyne e2v offers Space-grade digital components, including processors, memories, and processing modules, to support the Edge computing in Space. The LX2160-Space is a latest-generation Radiation-Tolerant processor featuring 16-cores ARM® Cortex®-A72, that offers a disruptive computing capability of 200,000 DMIPs / 280 GFLOPs, together with a large number of high-speed interfaces such as PCIe Gen3.0 and up to 100 Gb Ethernet to interface with other subsystems. These features make the LX2160-Space an ideal candidate for performing multiple AI tasks in flight, and to be the ”brain” of the spacecraft.
This white paper discusses the implementation of concurrent AI image processing using Klepsydra software running on the Teledyne e2v LX2160-Space processor. First, the general AI as a Service (AIaaS) approach is presented. Then, the scenario taken for the case study is described, including the details on the neural networks. Finally, the performance key results are presented.
If you wish to read more, please click on the following link to download the complete paper here.
Please fill out & submit the form to receive an answer from our Sales Team or Application Engineers.