Poster: Data Age Analyzer: Non-Intrusive Data Age Tracing for Real-Time ROS 2 Pipelines

Abstract

ROS 2 applications can satisfy per-callback latency targets while acting on stale data accumulated across multi-hop publish-subscribe chains. This poster presents Data Age Analyzer, a non-intrusive method that reconstructs message causal links from standard ros2_tracing LTTng traces, computes direct and indirect data age, and maps freshness misses to responsible roots, hops, and timing segments without changing nodes or message definitions. On 300 s Autoware Reference System traces, the analyzer shows that multi-threading reduces local executor queueing, with the worst values of medians of direct ages of topics dropping from 4.07 ms to 0.57 ms. However, >100 ms freshness misses are localized to a few root-to-leaf paths, showing that local hop improvement does not imply uniform root-to-leaf freshness.