Model activity-dependent cue admission in the posterior

Incorporate the admission mechanism into the posterior event-inference model when cue admission is statistically dependent on the physical activity event or on the cue realizations, rather than conditionally independent of them.

Background

The indoor activity-inference instantiation assumes that the admitted cue subset is conditionally independent of both the activity event and the cue realizations, given the wireless operating condition and control policy. Under this assumption, the admitted subset only determines which cue-likelihood factors enter the Bayesian posterior and contributes no additional information about the event itself.

The paper explicitly leaves unresolved the more general case in which the admission process depends on the activity being inferred or on the generated cue values. In that setting, the admission indicators would themselves carry event information and would need to be included in the posterior model, potentially changing the residual-uncertainty and Bayes-error calculations.

References

Admission is assumed conditionally independent of the activity event and cue realisations given $\boldsymbol{\xi}$ and $\bm{u}$, so $S_\Delta(t)$ determines which cue-likelihood factors enter the posterior without itself providing additional event information. If this assumption does not hold, the admission mechanism must also be included in the posterior model and is left for future work.

— Event-Inference Reliability for Physical AI over Wireless Networks  (2608.30663 - Mishra et al., 31 Aug 2026) in Section 5.2, “Wireless Evidence Delivery and TWI Admission” (immediately before Section 5.3)