Fault-Tolerant Dynamic Scheduling and Routing for TSN based In-vehicle Networks
Fault-Tolerant Dynamic Scheduling and Routing for TSN based In-vehicle Networks
复制标题
基于 TSN 的车载网络的容错动态调度和路由
DOI：
--
复制
发表时间：
2021
期刊：
影响因子：
--
通讯作者：
Madhu Chandra
中科院分区：
文献类型：
--
作者：
Ammad Ali Syed;S. Ayaz;T. Leinmüller;Madhu Chandra
功能安全被认为是未来自动驾驶汽车的最高要求之一，它通过车载网络（IVN）在传感器、执行器和控制器之间消耗大量的安全关键数据交换。 IVN应该提供足够的保证来满足此类自动驾驶汽车应用的高可靠性要求。为了避免数据包失败，时间敏感网络（TSN）在IEEE 802.1CB标准中提出了一种称为帧复制和可靠性消除（FRER）的容错机制。 FRER的主要思想是通过源和目的地之间的多条不相交的路径传输安全关键数据，这样如果一条路径无法传输安全关键数据，另一条路径仍然可以将数据包传送到目的地。本文分析了四种不同的动态调度和路由启发法，以支持基于 TSN 的 IVN 中的 FRER 功能。这些启发式方法使用组合分值，每个分值专用于单个路径，作为动态调度和路由多个冗余路径中的传入流的措施。其中一种算法，容错瓶颈启发式 (FTBH) 在可调度性和响应时间方面优于其他算法。与其他开发的启发式方法相比，它根据网络负载调度大约 3.0-7.0% 的流量。
Functional safety is considered as one of the utmost requirements of the future autonomous vehicle, which consumes an ample amount of safety-critical data exchange among sensors, actuators and controllers through the in-vehicle network (IVN). IVN should provide enough guarantees to satisfy the high-reliability requirement of such autonomous vehicle applications. To avoid packet failures, Time-sensitive Network (TSN) proposes a fault-tolerant mechanism called frame replication and elimination for reliability (FRER) in IEEE 802.1CB standard. The main idea of FRER is to transmit safety-critical data via multiple disjoint paths between source and destination so that if one path fails to transmit the safety-critical data, the other path can still deliver the packet to its destination. In this paper, four different dynamic scheduling and routing heuristics are analyzed to support FRER functionality in TSN based IVN. These heuristics use combined score value, each score value is dedicated for a single path, as a measure to schedule and route incoming flow in multiple redundant paths on the fly. One of the algorithms, the Fault-Tolerant Bottleneck Heuristic (FTBH) outperforms others in terms of schedulability and response time. It schedules around 3.0–7.0% more flows as compared to other developed heuristics depending on the network load.