Modeling and Scheduling of Fusion Patterns in Autonomous Driving Systems (Extended Version)
Hoora Sobhani, Hyoseung Kim
TL;DR
The paper addresses the problem that state-of-the-art ADS DAG scheduling often oversimplifies data fusion by assuming fixed triggering patterns. It introduces an ILP-based framework that models five fusion patterns (sen, sub, t-fus, w-fus, i-fus) and branch-then-fusion paths, generating deterministic offline schedules across multi-core platforms using producer and predecessor structures with binary coupling variables. The framework optimizes multiple real-time metrics, including $MRT$, $MTD$, $PAoI$, $WCRT$, and $MS$, by formulating fusion-specific constraints and end-to-end mappings between sensor and actuator instances, with analysis performed over a horizon $\Delta = 3\cdot HP$. Evaluation against the state-of-the-art and on randomly generated DAGs demonstrates improved handling of diverse fusion patterns and meaningful gains in timing performance, with practical validation on a Raspberry Pi and broad applicability to ADS software stacks. This work provides a configurable tool for ADS designers to compare fusion strategies, quantify trade-offs, and derive offline schedules that can be deployed on real platforms, potentially influencing sensor fusion design and scheduling policies in safety-critical autonomous driving systems.
Abstract
In Autonomous Driving Systems (ADS), Directed Acyclic Graphs (DAGs) are widely used to model complex data dependencies and inter-task communication. However, existing DAG scheduling approaches oversimplify data fusion tasks by assuming fixed triggering mechanisms, failing to capture the diverse fusion patterns found in real-world ADS software stacks. In this paper, we propose a systematic framework for analyzing various fusion patterns and their performance implications in ADS. Our framework models three distinct fusion task types: timer-triggered, wait-for-all, and immediate fusion, which comprehensively represent real-world fusion behaviors. Our Integer Linear Programming (ILP)-based approach enables an optimization of multiple real-time performance metrics, including reaction time, time disparity, age of information, and response time, while generating deterministic offline schedules directly applicable to real platforms. Evaluation using real-world ADS case studies, Raspberry Pi implementation, and randomly generated DAGs demonstrates that our framework handles diverse fusion patterns beyond the scope of existing work, and achieves substantial performance improvements in comparable scenarios.