EnCost: uplink-cost-aware cluster-head selection with sink-trajectory look-ahead for energy-efficient mobile-sink wireless sensor networks
Year: 2026
Type: article
Abstract: In a mobile-sink wireless sensor network (WSN), the cluster-head (CH)-to-sink uplink periodically enters the distance-fourth-power multipath regime, where electing a CH far from the sink incurs a disproportionate energy penalty. We propose EnCost-CR , a closed-form CH-selection rule that scores each node by its residual energy minus the modelled energy of the uplink it would pay as a head, instead of the normalized distance used by existing energy- and distance-aware rules. We are explicit that this is cost accounting carried out before the decision, not prediction: it uses the sink position already observed for the round. EnCost-CR is energy-primary, preserving rotation and load balance, and runs in linear time with no per-round search. Across 30 Monte-Carlo deployments of 100 nodes under a mobile sink it attains the longest first-node-death stability period of every protocol in our main comparison, exceeding LEACH, DEEC and HEED by 24–42 percent, at 54–107 times lower measured per-round cost than the population searches, and it equals an offline oracle on a static sink; charging the whole head cost, not the uplink alone, improves it slightly further. Given the same physical energy objective, however, a per-round metaheuristic exceeds it, so the claim we make is a cost-adjusted one: a comparable stability period for a single O ( N ) pass, not a lifetime that search cannot reach. Measuring delivery at the source level under explicit aggregation semantics, not by counting aggregate packets, reverses an apparent throughput advantage over HEED. A genuinely predictive variant, driven by a causal online forecast of the sink track that is accurate to a few percent of its excursion, does not improve the stability period at any setting tested, which is direct evidence that anticipation is not the operative mechanism. We map the rule’s boundary conditions explicitly: it requires the uplink cost to remain small relative to the spread of residual energies, and it falls below plain energy balancing when the sink position is poorly known. Because the best rule is unknown before deployment, we add a sliding-window upper-confidence-bound bandit hyper-heuristic; we show that its usual regret guarantee does not transfer, because the arm rewards are state-dependent, and report measured dynamic regret instead. The complete MATLAB implementation, which regenerates every table, figure and quoted number, is released.
Source: Scientific Reports
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7221019712