On-Orbit Remote Sensing Image Processing Complex Task Scheduling Model Based on Heterogeneous Multiprocessor
Qiangqiang Jiang,Haipeng Wang,Qinglei Kong,Yamin Zhang,Bo Chen
DOI: https://doi.org/10.1109/tgrs.2023.3327279
IF: 8.2
2023-01-01
IEEE Transactions on Geoscience and Remote Sensing
Abstract:Nowadays, the proliferation of small satellites brings the skyrocketing rise in space data, especially the shift to on-orbit computing needs. On one hand, with the increasing volume of data generation, like high-resolution remote sensing images, on-orbit computing produces near real-time onboard solutions and quick responses. However, constrained by the limited size and energy supply of satellites, achieving energy-efficient on-orbit computing remains a crucial challenge. In this article, an on-orbit remote sensing image processing complex task scheduling model facing a heterogeneous multiprocessor system (HMPS) is proposed. First, aiming at accelerating image processing, we establish a novel parallel task execution model using a directed acyclic graph (DAG) to universally describe typical missions, i.e., cloud detection, geometric correction, and image classification. Subsequently, a mathematical task scheduling formulation is defined to calculate the makespan, and total energy consumption (TEC) required when executing DAG on HMPS. Second, a new Pareto-based iterated greedy optimizer (PIGO) is devised to complete the energy- and time-efficient task execution and resource allocation on HMPS through confined inserting mutation, destruction-reconstruction, and local search. Finally, we build an emulated on-orbit HMPS to conduct experiments. The results show that, in comparison with the scheme without model scheduling, the cost savings of around 51% and 54% in makespan and TEC, respectively, are achieved by the proposed model. Moreover, the HMPS configured with our methodology can obtain $2.2{\times }$ improvement in energy efficiency and process up to $2.56\times 10^{5}$ pixels per unit of power (W) and time (s).