R3TOS-Based Integrated Modular Space Avionics for On-Board Real-Time Data Processing

摘要

The partitioning of computing platforms is a wellknown technique for achieving fault isolation and fault tolerance in space avionics. With the advent of large-capacity and partially reconfigurable FPGAs, partitioning has been proposed at the System-on-Chip (SoC) level. However, fixed hardware partitions in a SoC, such as those in slotted reconfigurable architectures, are not effective in coping with permanent chip damage provoked by cumulative space radiation, the reason being that a single damaged on-chip resource in one partition can prevent the use of that entire partition. In this paper, we discuss how to use R3TOS (Reliable Reconfigurable Real-Time Operating System) to implement a dynamically-partitioned computing platform on a Xilinx Zynq SoC that processes science data delivered by a NASA Fourier transform spectrometer. Unlike in related approaches, R3TOS avoids fixed partitions by harnessing the Zynq's configuration memory for exchanging data among the partitions. This enables us to achieve higher levels of flexibility in the spectrometer avionics, which results in a more effective capability to withstand chip damage. We have simulated different chip damage scenarios and checked that an R3TOS-based prototypic avionics system can tolerate on average around 13% more on-chip damaged resources than a traditional fixed slotted solution.