VIOVIO Survey BenchmarkGitHub Pages reproducible benchmark

Lesson 09 / Advanced

Advanced estimator assumptions

A deeper structure for experts reviewing estimator scope, state, initialization, calibration, and active-state maintenance.

Learning outcomes

  • Separate odometry scope from broader system scope
  • Ask what state is estimated and what remains fixed
  • Review initialization and calibration assumptions before interpreting results

Estimator scope

The public benchmark reports VIO / odometry output.

Systems with broader modules must be configured so the reported trajectory is the odometry trajectory, not a future-corrected publication artifact.

Shared review schema

For each system, experts should ask about core state, active-state maintenance, IMU handling, visual residuals, update mechanism, initialization, and calibration assumptions.

This mirrors the report schema without forcing every method into one implementation style.

Checklist

  • State variables are documented
  • Fixed calibration assumptions are documented
  • Initialization assumptions are documented
  • Output scope is documented

Advanced notes

Active-state maintenance is the general concept: clones, keyframes, fixed windows, local map nodes, or other retained temporal state are all specific cases.

Calibration mismatch and timing choices can dominate benchmark outcomes even when the estimator backend is sound.

Practice task

Pick one public system and write four lines: scope, state, initialization, calibration assumptions.