Fast Real-Time Scheduling for Ethernet-Based Train Control Networks.

Qinghan Yu,Tian Wang,Xibin Zhao,Hai Wan,Yue Gao,Chenyang Lu,Ming Gu
DOI: https://doi.org/10.1109/bdcloud.2018.00085
2018-01-01
Abstract:Traditional field bus cannot meet the demand of modern trains that generate increasing amounts of control and diagnostic data. To support these communications with real-time requirements, time-triggered network is a promising technology as real-time frames are transmitted according to a pre-computed schedule. However, a unique challenge for modern trains is dynamic train inauguration, where the pre-computed schedule needs to be reconfigured online due to network topology changes. This paper presents a novel scheduling approach specifically designed for the dynamic schedule generation. The novelty of our approach is two-fold. In contrast to the traditional scheduling approaches that rely on time-consuming solvers to generate schedules offline, our scheduler features efficient heuristics-based algorithms. Moreover, our scheduling algorithm is cognizant of the interdependence between bandwidth and memory constraints in time-triggered network switches with shared buffers. Simulation results show that our scheduling algorithm has better scheduling ability than the myopic approach and SMT-based approach when schedules are required to be generated rapidly for train inauguration.
What problem does this paper attempt to address?