Kalman filter benchmarks #1407
Kalman filter benchmarks #1407
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| ; predict_x | ||
| (let ([x0* (+ x0 (+ x1 (* 0.5 (* dt (* dt x2)))))] | ||
| [x1* (+ x1 x2)] | ||
| [x2* x2]) |
There was a problem hiding this comment.
The predicted state vector does not use the transition matrix coefficients that include dt. x0* adds x1 directly and x1* adds x2 directly, so the position and velocity are advanced as if the time step were always 1 second. For any other dt the predicted state is wrong and Kalman updates use inconsistent P and x propagation. Multiply the velocity by dt and acceleration by dt in these assignments (e.g. (+ x0 (* dt x1) …) and (+ x1 (* dt x2))).
Useful? React with 👍 / 👎.
There was a problem hiding this comment.
That's true, but still, nice to keep them to preserve a complete behavior of the initial program
This PR adds new benchmarks.
These benchmarks represent Kalman filter implementation that is taken from: Elaeina repository.
Kalman filter benchmarks are parsed per each output variable.
Despite that the benchmarks share a large part of executions - they differ in an output variable.