This is a 2D-native particle filter (sequential Monte Carlo localization) dashboard — the same map/dot canvas that a real navigation engineer would actually plot, rather than a 3D scene viewed from above. A cloud of N pose hypotheses (particles) is propagated and re-weighted every step:
Predict: x_i ← x_i + Δu + N(0, σ_drift²) (odometry + slip noise)
Update: w_i ← exp( -(z - h(x_i))² / (2·σ_sensor²) ) (Bayes likelihood)
Estimate: x̂ = Σ w_i·x_i / Σ w_i (weighted mean)
Health: N_eff = 1 / Σ w_i² (effective sample size)
Resample: when N_eff/N falls below the threshold, redraw particles proportional to w_i (systematic resampling)
- Field map (top panel) — a flat heatmap of the magnetic anomaly field, drawn directly in 2D (no camera or perspective). Drag to pan, scroll/pinch to zoom.
- Weight histogram (middle panel) — a live bar chart of the particle weight distribution. A single tall spike means the filter is confident; a flat spread means it is lost — information the 3D orbit view cannot show at all.
- ESS strip chart (bottom panel) — a scrolling trace of N_eff/N over time, so you can see exactly when a resampling event fires (the threshold line).
Informativeness comes from the field's spatial variability, not its raw strength: the cloud tightens fastest crossing a steep anomaly gradient and drifts loosest over flat regions, exactly as it does for real magnetic-anomaly navigation used in submarines and GPS-denied aircraft.