Dynamic Scheduling and Routing for TSN based In-vehicle Networks

The future autonomous vehicle is not only processing the copious amount of indispensable data generated by its onboard sensors but also utilizing the data from other vehicles, roadside unit (RSU) etc. Managing the mixed-criticality data requires intelligent time-sensitive scheduling and routing within the in-vehicle network (IVN) infrastructure. Use-cases related to self-adaptivity (including vehicular communication), partial networking and embedded virtualization require to change the configuration of the IVN at runtime. State-of-the-art IEEE Time-Sensitive Networking (TSN) standards possess a grave challenge in handling runtime reconfigurations. Above mentioned use-cases foster the development of scalable and efficient dynamic scheduling and routing algorithms for TSN based IVN. In this paper, four meticulously designed heuristics are analyzed for dynamic scheduling and routing on-the-fly in TSN based IVN. One of the algorithms, Bottleneck heuristic outperforms others in term of schedulability and response time. It schedules around 16 − 22% more flows as compared to other developed heuristics depending on the network load.

I. Introduction

Advancements in the development of various on-board sensor technologies including camera, lidar and radar as well as vehicular communication technologies pave the way for the development of the autonomous vehicle. Different electronic components installed in the vehicle rely on the in-vehicle network to communicate with each other. The demand for electric and electronic (EE) systems is expanding exponentially with the increasing number of in-vehicle applications. Modern vehicle may implement new applications by adding new electronic control unit (ECU) as well as new actuators/sensors in the vehicle network. Some state of the art in-vehicle communication technologies include controller area network (CAN), FlexRay, local interconnect network (LIN), Media Oriented Systems Transport (MOST). As the automation level of vehicle increases, it will require more intelligence supported by a diverse set of actuators/sensors and ECUs to fulfil the functional safety requirements. These actuators/sensors are connected through the in-vehicle network on the physical level and interacting with the central controller (i.e., central processing unit (CPU)). These sensors will generate an ample amount of data which will traverse the IVN to reach to CPU which performs different tasks (e.g., sensor fusion, driving monitoring, motion/mission planning, driving control) based on the received information. The low bandwidth legacy IVN technologies cannot support the ample amount of data required by new applications. The gradual increase in the number of ECUs and the complexity in the IVN lead to demand for more bandwidth and scalable communication technologies. Therefore, Ethernet and its Time-Sensitive Networking (TSN) extensions are now being considered as a potential candidate for next-generation vehicles because of its low cost, scalable bandwidth, flexibility and advanced capabilities supporting mixed critical in-vehicle services (i.e., time-triggered, rate constraint, and best-effort) with different data rates.