Hierarchical Residual-Envelope Trust: A Deterministic Framework for Grouped Multi-Sensor Fusion

Hierarchical Residual-Envelope Trust (HRET) extends the Drift--Slew Fusion Bootstrap (DSFB) framework to structured multi-sensor systems with explicit sensor grouping and correlated disturbance modeling. Unlike stochastic or covariance-based approaches, HRET operates under deterministic bounded disturbance assumptions and modulates trust through residual-envelope dynamics at both channel and group levels. Each sensor channel maintains a residual envelope governing channel-level trust, while group envelopes aggregate residual magnitudes across predefined sensor clusters to capture correlated degradation. Trust weights are composed multiplicatively and normalized to preserve strict convexity, ensuring positive weights and unity sum at all times. The framework guarantees bounded envelope dynamics under bounded disturbances and enables coordinated suppression and recovery of sensor groups under correlated corruption. HRET reduces to DSFB as a special case when each channel forms its own singleton group. The formulation is fully deterministic, convexity-preserving, and implementation-aligned with the existing DSFB architecture. The framework assumes bounded disturbances and predefined grouping, and does not claim statistical optimality.