Robust Deadlock Avoidance for Dependent-Task Offloading in Remote-Sensing Satellite–Terrestrial Networks with an Unreliable Link and Buffer Constraints
Year: 2026
Type: article
Abstract: Remote-sensing satellite–terrestrial networks (RSTNs) with mobile edge computing (MEC) support the processing of Earth-observation data submitted by ground terminals. However, when these applications comprise multiple subtasks with strict data dependencies, their concurrent execution over satellites with limited buffers and an unreliable inter-satellite link (ISL) introduces severe risks of system deadlocks and task non-completion. A robust deadlock avoidance policy (robust DAP) for dependent task offloading in RSTNs with an unreliable ISL and limited buffers (RSTN-UILB) is developed in this study to address this issue. Specifically, a Petri net is constructed to characterize task dependencies, buffer constraints, and autonomous ISL disruption and recovery. Based on this model, robust-completable states and robust DAPs are defined, and a necessary and sufficient condition for preserving a permitted task-only completion path from every controlled reachable state is established. A robust modified banker’s algorithm (RMBA) is designed to certify robust completability. Leveraging the RMBA, a robust DAP is synthesized to preserve robust nonblocking with initial RMBA certification, with polynomial evaluation complexity. Extensive simulations show near-optimal mean robust permissiveness on fully enumerated test instances at substantially lower average certification cost than exhaustive search. Experiments demonstrate broader observed permission than conservative admission and completion across diverse network scales and heterogeneous buffers.
Source: Sensors
DOI: 10.3390/s26227319
Authors: Wenji Yang, Yikang Yang, Chao Fang, Yanxiang Feng
Institution: Xi'an Jiaotong University
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $3,137
OpenAlex ID: W7236608454
