Can Linux Fly a Drone?: A Flight Control Evaluation for Unified UAV Architectures Using PREEMPT_RT KernelShow others and affiliations
2026 (English)In: 2026 IEEE 31st International Conference on Emerging Technologies and Factory Automation (ETFA), Institute of Electrical and Electronics Engineers (IEEE) , 2026, p. 1-8Conference paper, Published paper (Refereed)
Abstract [en]
Modern UAV architectures increasingly aim to unify high-level autonomy and low-level flight control on a single General-Purpose Operating System, but complex multicore System-on-Chips introduce significant timing indeterminism. This paper performs an architectural schedulability analysis of the PREEMPT_RT Linux kernel on Raspberry Pi 5, isolating the impact of kernel activation paths (deferred SoftIRQs versus realtime direct activation) on a 250-Hz control loop, and evaluates the closed-loop consequences through a Software-in-the-Loop experiment. Under heavy stress, the standard kernel exhibits worst-case latencies above 9 ms, and even POSIX real-time policies spike to 1.8 ms. PREEMPT_RT's direct wake-up path reduces this by 88% to under $225 \mu ~\mathrm{s}$. The SIL experiment indicates the operational consequences: the standard kernel configuration loses stability and crashes, while PREEMPT_RT completes the reference trajectory. While PREEMPT_RT bounds OS-level latency, the residual jitter on modern edge-AI SoCs is primarily driven by shared memory hierarchy contention.
Place, publisher, year, edition, pages
Institute of Electrical and Electronics Engineers (IEEE) , 2026. p. 1-8
Series
IEEE Conference on Emerging Technologies and Factory Automation, ISSN 1946-0740, E-ISSN 1946-0759
Keywords [en]
Kernel, Timing, Real-time systems, Standards, Linux, Human factors, Stress, Jitter, Scheduling, Printing, UAV Flight Control, Real-Time Linux, PREEMPT_RT, Embedded Systems, Real-Time Scheduling, Raspberry Pi
National Category
Control Engineering
Identifiers
URN: urn:nbn:se:mdh:diva-79391DOI: 10.1109/ETFA68377.2026.11724768OAI: oai:DiVA.org:mdh-79391DiVA, id: diva2:2114852
Conference
2026 IEEE 31st International Conference on Emerging Technologies and Factory Automation (ETFA), 8-11 September, 2026, Västerås, Sweden
2026-11-032026-11-032026-11-03Bibliographically approved