Work

Cross-Layer Framework for Fast Steiner Tree Optimization With RAFT Commit-Time Guarantees in HAPS Meshes

Year: 2026

Type: article

Abstract: While High Altitude Platform Station (HAPS) mesh topologies are quasi-static over a service interval—owing to platform station-keeping—the quality of individual links may still vary dynamically. We investigate the cross-layer impact of running RAFT, a widely adopted consensus protocol that requires a majority quorum for decision enforcement, over HAPS networks. Specifically, we focus on the commit time $T_{c}$ , defined as the time required to obtain consensus, which must satisfy the latency bound stipulated in the service level agreement (SLA). Because HAPS platforms combine station-keeping with directional antennas that track slowly moving peers, they can sustain a given topology over extended periods. Given such a HAPS topology, we associate $T_{c}$ with a Steiner Tree Problem, formulate it as a commit latency constraint (STPG- $T_{c}$ ), and present a mixed-integer nonlinear programming (MINLP) model that jointly captures network dynamics and RAFT timing semantics. In the resulting design, a lookup table (LUT) delivers millisecond-scale feasibility checks by exploiting path-score monotonicity, boundary analysis, and stripe-based space decomposition. Building on this LUT, we develop the Dynamic Tempering Algorithm (DTA) as an efficient STPG- $T_{c}$ solver. Across our test cases the solver consistently returns optimal or near-optimal solutions, whereas a commercial MIP solver (Gurobi) fails to reach optimality in 40% of cases within a 5-second limit. Our method reduces computation time by more than a factor of twenty while improving accuracy, thereby enabling scalable and SLA-aware consensus in a dynamic HAPS environment.

Source: IEEE Access

DOI: 10.1109/access.2026.3738422

Authors: Yuqiang Wen, K. L. Eddie Law

Institution: Macao Polytechnic University

Language: English

Topic: Satellite Communication Systems

Subfield: Aerospace Engineering

Field: Engineering

Domain: Physical Sciences

SDG: Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $2,075