Why Mathematics? | Covariance Intersection for Reliability Sensor Fusion: Unknown Correlations, Information Weights, Consistency and Optimisation

Trace where each estimate obtained its evidence before fusing it. Shared priors, common process noise, relayed measurements and overlapping training data create correlation between estimation errors. Two maintenance agents both used the same inspection report. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Treating them as independent counts evidence twice. Map information provenance.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make why unknown correlation matters the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Did You Know? Covariance intersection performs a convex combination in information space, so the chosen weight controls both the fused centre and its uncertainty. A small observation like this often reveals the conceptual boundary that a formula alone hides.

Contents · Next chapter

Information weights combine precision while protecting against unknown overlap. A covariance is consistent when it does not understate the actual estimation error covariance under the stated framework. A reliable interval should not be narrower than the information supports. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Consistency is not the same as unbiasedness or accuracy. Evaluate coverage and error.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make consistency the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

The conservative covariance records what correlation knowledge is missing. CI combines inverse covariances using nonnegative weights that sum to one. Scalar precisions one quarter and one are blended. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Singular covariances need specialised handling. Check definiteness and rank.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make information-space fusion the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

Optimisation chooses a matrix criterion, not a vague preference. At weight one half, compute 0.5 times 0.25 plus 0.5 times 1. The fused precision is 0.625. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Averaging variances directly is a different operation. Write the information equation first.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make worked fused precision the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Primary research source: https://doi.org/10.1109/ACC.1997.609105. Read its problem definition, mathematical criterion and assumptions before transferring the method to a reliability setting.

Contents · Previous chapter · Next chapter

Trace where each estimate obtained its evidence before fusing it. Invert the fused scalar precision. One divided by 0.625 gives 1.6. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Matrix inversion should normally be implemented with stable solves. Use factorisations in software.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make worked fused variance the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Did You Know? Covariance intersection performs a convex combination in information space, so the chosen weight controls both the fused centre and its uncertainty. A small observation like this often reveals the conceptual boundary that a formula alone hides.

Contents · Previous chapter · Next chapter

Information weights combine precision while protecting against unknown overlap. Combine information-weighted means and multiply by the fused variance. The result is 10.4. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. A simple arithmetic mean would ignore unequal uncertainty. Keep every weight visible.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make worked fused mean the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

The conservative covariance records what correlation knowledge is missing. Independent Gaussian fusion adds full precisions and gives variance 0.8 in the teaching case. CI is wider because dependence is unknown. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. CI should not replace exact fusion when cross covariance is known. Use the strongest justified model.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make comparison with independence the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

Optimisation chooses a matrix criterion, not a vague preference. Choose ω to minimise a scalar measure of fused covariance, commonly determinant or trace, over the unit interval. A one-dimensional line search finds the preferred compromise. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Different objectives can choose different weights. State the objective and optimiser.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make weight optimisation the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

Trace where each estimate obtained its evidence before fusing it. Minimising log determinant targets uncertainty-volume reduction and improves numerical scaling. A multivariate ellipsoid becomes smaller in volume. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Volume can hide one poorly estimated direction. Inspect eigenvalues too.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make determinant objective the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Did You Know? Covariance intersection performs a convex combination in information space, so the chosen weight controls both the fused centre and its uncertainty. A small observation like this often reveals the conceptual boundary that a formula alone hides.

Contents · Previous chapter · Next chapter

Information weights combine precision while protecting against unknown overlap. Minimising trace targets the sum of marginal variances. Average coordinate uncertainty is reduced. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Trace depends on state scaling and units. Nondimensionalise or justify weights.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make trace objective the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

The conservative covariance records what correlation knowledge is missing. CI selects an uncertainty ellipsoid consistent with both input descriptions under unknown correlation. Two elongated ellipses constrain different directions. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. A visually small ellipse is not automatically calibrated. Test empirical coverage.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make geometry of ellipsoids the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

Optimisation chooses a matrix criterion, not a vague preference. CI can be extended or applied sequentially to more than two estimates. Three nodes contribute local health estimates. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Sequential pairwise fusion can depend on order. Document architecture and test permutations.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make multiple estimates the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Supporting authoritative source: https://ieeexplore.ieee.org/document/609105. Compare the published scope with this teaching example, especially the information set, numerical representation and conditions under which the result is claimed.

Contents · Previous chapter · Next chapter

Trace where each estimate obtained its evidence before fusing it. Nodes can fuse neighbour estimates without tracking every cross-correlation path. Packet routing creates unknown information overlap. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Communication delay and stale data remain separate problems. Timestamp every message.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make distributed networks the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Did You Know? Covariance intersection performs a convex combination in information space, so the chosen weight controls both the fused centre and its uncertainty. A small observation like this often reveals the conceptual boundary that a formula alone hides.

Contents · Previous chapter · Next chapter

Information weights combine precision while protecting against unknown overlap. If shared information can be identified, explicit subtraction or track-to-track methods may be less conservative. A common prior is factored out before fusion. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Incorrectly removing common information can become overconfident. Use provenance and audits.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make common information the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

The conservative covariance records what correlation knowledge is missing. CI can fuse Gaussian approximations produced by nonlinear filters, but the input covariances must still be meaningful. Two UKF nodes report state and covariance. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Gaussian summaries may hide multimodality. Inspect input distributions.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make nonlinear estimates the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

Optimisation chooses a matrix criterion, not a vague preference. CI protects against unknown cross correlation at the cost of potentially wider uncertainty. A safety threshold is evaluated with a cautious interval. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Excessive conservatism can reduce useful sensitivity. Compare with bounded-correlation methods.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make robustness and conservatism the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

Trace where each estimate obtained its evidence before fusing it. Simulate varying cross correlations and check empirical error against reported covariance. Coverage is tested from zero to strong dependence. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Testing only independent cases misses the purpose. Stress the full plausible range.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make validation the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Did You Know? Covariance intersection performs a convex combination in information space, so the chosen weight controls both the fused centre and its uncertainty. A small observation like this often reveals the conceptual boundary that a formula alone hides.

Contents · Previous chapter · Next chapter

Information weights combine precision while protecting against unknown overlap. Record input estimates, covariances, provenance, objective, optimal weight, numerical method and final result. A reviewer can reproduce the scalar example. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Saying only fused by CI is incomplete. Publish the audit trail.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make reporting the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Contents · Previous chapter · Next chapter

The conservative covariance records what correlation knowledge is missing. The example connects reciprocals, weighted averages, variance and cautious reasoning. A learner can compare variance 1.6 with 0.8 and explain why. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. Wider uncertainty is not mathematical failure. Link width to assumptions.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make student and parent guidance the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Useful next reading: distributed Kalman filtering, Kalman filtering foundations, Mathematics Learning Hub, eduKate Mathematics Learning System, approach to learning mathematics, Bukit Timah Mathematics curriculum overview, cross-subject learning approach. Follow the link that answers the next question in your own reasoning, rather than treating the list as a reading quota.

Contents · Previous chapter · Next chapter

Optimisation chooses a matrix criterion, not a vague preference. Covariance intersection is for unknown cross correlation, not a universal replacement for probabilistically exact fusion. Vary ω and plot the fused mean and variance. The mathematics is useful only when the state, units, timing and information boundary stay attached to every number. The weight is not automatically a trust score. State what the optimisation actually minimises.

A careful audit has four moves. First write the claim in ordinary language. Next preserve the intermediate quantities instead of jumping to the final estimate. Then check dimensions, signs, matrix validity or probability totals. Finally compare the result with the physical story and an alternative model. Save input means and covariances, provenance, correlation knowledge, objective, search interval, chosen weight, fused information and empirical coverage. This record lets another reader locate whether disagreement comes from data, assumptions, computation or interpretation.

For practice, make practice, misconceptions and next reading the focus and use weight one half to combine precisions one quarter and one, obtaining fused precision zero point six two five, variance one point six and mean ten point four. Predict the direction before using a calculator. Change one defensible input, recompute the result and explain why it moved. Finish with one supported conclusion, one uncertainty and one next check. That routine develops mathematical problem-solving skills that transfer beyond this one algorithm.

Quick answers

Why is mathematics important here? It turns a verbal concern about uncertainty, later evidence, shared information or robustness into quantities that can be checked and challenged.

Does a successful calculation prove the physical model? No. It shows what follows from the declared model, inputs and criterion. Validation against independent evidence remains necessary.

Which school mathematics transfers? Ratios, inequalities, functions, weighted averages, variance, graphs and matrices all reappear, joined by careful language about time and assumptions.

How should a student practise? Reproduce the small example, keep intermediate values, predict the direction of change and explain one limitation without jargon.

What should a practitioner report? Equations, data boundaries, initialization, uncertainty settings, numerical method, diagnostics, sensitivity tests and the rule that connects estimates to decisions.

How can a parent help? Ask what each number represents, when the information became available and what assumption would most change the conclusion. Explanation is part of mathematical mastery.

Contents · Previous chapter · Mathematics Learning Hub