Journal article
Ensuring schedulability and end-to-end latency in DAG-based tasks for AUTOSAR
Journal of systems architecture, Vol.178, 103893
09/2026
DOI: 10.1016/j.sysarc.2026.103893
Abstract
AUTOSAR (AUTomotive Open System ARchitecture) is a standardized software framework widely adopted in multi-core autonomous driving systems. Compared with traditional real-time systems that mainly face schedulability and end-to-end latency challenges, AUTOSAR introduces additional complexities due to its characteristics, including the SWC (Software Component) architecture, constraints on task assignment, core binding requirements, and others. In this paper, we propose a three-stage static scheduling method to satisfy both schedulability and end-to-end latency for AUTOSAR. First, a multi-rate directed acyclic graph (DAG) model is constructed based on the SWC architecture. Then, considering the constraints on task assignment, core binding requirements, and load balancing, tasks are allocated to the corresponding cores. During schedule generation, the proposed method performs scheduling in three stages: the task assignment stage, which allocates tasks to cores respecting SWC co-location and load balancing; the intra-SWC scheduling stage, which determines the execution order of jobs within each SWC; and the inter-SWC scheduling stage, which constructs the final static schedule. Experimental evaluations show that the proposed method efficiently generates static schedules under the AUTOSAR framework, ensuring both schedulability and end-to-end latency guarantees. Compared with baseline methods, schedulability is improved by up to 8%, while end-to-end latency is reduced by up to 20.9%.
•A DAG-based ILP method for scheduling multi-rate AUTOSAR components.•Ensures strict end-to-end latency and overall task schedulability.•Optimizes multi-core task assignment and job execution order.
Details
- Title: Subtitle
- Ensuring schedulability and end-to-end latency in DAG-based tasks for AUTOSAR
- Creators
- Zhiyong Wang - Northeastern UniversityFeng Li - Dalian University of TechnologyQingxu Deng - Northeastern UniversityTianyu Zhang - University of IowaYe Ma - Dalian University of Technology
- Resource Type
- Journal article
- Publication Details
- Journal of systems architecture, Vol.178, 103893
- DOI
- 10.1016/j.sysarc.2026.103893
- ISSN
- 1383-7621
- Publisher
- Elsevier B.V
- Language
- English
- Date published
- 09/2026
- Academic Unit
- Computer Science
- Record Identifier
- 9985175466802771
Metrics
1 Record Views