Latency and jitter are different stresses
Test steady delay and variable delivery timing separately.
Research note · November 14, 2026Research question
Can a controller tolerate average latency while failing under timing variation?
Method
- Begin with an ordered input stream whose event timestamps are preserved. Add a fixed transport delay in one case and a declared varying delay schedule in another.
- Keep the average delay comparable while recording the full distribution and worst observed delay. Define whether stale inputs are dropped, held or consumed when they arrive.
- Compare response quality and deadline misses under identical initial conditions. Preserve both source timestamps and arrival timestamps so replay can reproduce the timing policy.
Records to retain
- Source and arrival timestamps
- Delay schedule
- Stale-input handling policy
- State and deadline traces
Source basis
OpenTelemetry’s separation of event and observed time provides a useful recording model, while ROS time design clarifies clock domains. Average delay alone cannot describe a stream’s delivery order or deadline behavior.
Interpretation & limits
A network-delay fixture does not represent every physical sensor or bus. Synthetic jitter should not be presented as a measured deployment distribution.
Next experiments
- Construct a brief burst of delay with the same long-run mean.
- Test whether delayed data is visibly identified at the consumer boundary.
References
Technical reference
Implementation details supporting this note, drawn directly from the maintained documentation. These are technical references, not additional papers or new experimental results. Expand a topic to inspect its method, fields and limits.