A Hybrid Deterministic-Stochastic Cross-Layer Energy Model for Hierarchical Wireless Sensor Networks

Accurate current-consumption estimation in IEEE 802.15.4 Wireless Sensor Networks requires linking network operation to the hardware states that determine node energy use. This paper proposes a cross-layer energy-modeling methodology that combines experimentally characterized Current Consumption Profiles (CCPs), protocol-derived state-duration models, topology-dependent workload, and a two-stage probabilistic model for retransmissions. The methodology distinguishes structurally constrained states, empirically modeled state durations, and discrete stochastic communication events. In the evaluated CC2538–ContikiMAC implementation, transmission duration is modeled from payload size, ContikiMAC strobe activity, and Clear Channel Assessment (CCA) timing, while sleep duration is obtained from the residual communication-cycle constraint. Retransmission occurrence is represented by a Ridge-regularized logistic model using measured CCA duration, whereas the positive retry count is represented by its calibrated conditional mean. The resulting probabilistic component is integrated with the deterministic CCP formulation to estimate the additional charge associated with retry-enabled communication events. Full-network verification across 30 node–condition observations yielded an overall MAPE of 10.81%, WMAPE of 10.51%, and condition-specific MAPEs ranging from 4.60% to 15.78%.