I have a system that takes in many sensor inputs (~100 channels) and outputs a low-dimensional state estimate (~2 states). Traditionally, a Kalman Filter works well to smooth the noisy input data. However, the inputs here are asynchronous, i.e. sensor updates arrive at different times, with update intervals varying over time. Also, the sensor updates are relatively infrequent.

The system currently performs a regression/state-update at constant time intervals using the most recent sensor values, however, this means some channels will be regressed using "old" data. I'm wondering if there's some version of the Kalman Filter or a similar regression model that can better handle these asynchronous updates, perhaps running state-updates at a non-constant rate?

  • $\begingroup$ You can model this either as missing data at a regular time step (the finest one), or with time-varying dynamics depending on the step size. The Kalman filter can already handle either. $\endgroup$ – Chris Haug May 1 at 1:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.