miss_rate

Formula

miss_rate = count(samples where trajectory_fde > threshold) / sample_count

Default threshold: 2.0.

Hand-Computable Example

Sample FDE values: 2.0 and 0.5, threshold 1.0.

Expected output: 0.5.

Edge Cases

  • Empty sample list returns 0.0.
  • The comparison is strict: FDE equal to threshold is not a miss.

Limitations

Miss rate depends heavily on the selected threshold and task horizon.

Test References

  • tests/golden/test_openloop_metrics_golden.py
  • tests/test_openloop_metrics.py