Time-Predictable Ethernet Communication in Automotive: What are the Challenges?

Scientific Conference (2023, March 22)

Software defined vehicle puts the challenges not only on the computing systems in the vehicle but also on the network. To design a system that can exhibit the required level of determinism for a set of ... [more ▼]

Software defined vehicle puts the challenges not only on the computing systems in the vehicle but also on the network. To design a system that can exhibit the required level of determinism for a set of distributed control applications, it is not enough to have a network that is predictable from the timing perspective, one also needs proper real-time scheduling mechanisms in the operating system and in the communication stack, i.e. the interface between the application and the network. In this contribution, we present: - Overall concept, or what do we mean with a time-predictable system, - State of the art in deterministic real-time Ethernet communication, - The real-time scheduling mechanisms commonly used in Automotive, - Examples of challenging issues and practical use cases that OEMs need to address toward timing predictability in the system. [less ▲]

Approximating WCRT through the aggregation of short simulations with different initial conditions: application to TSN

in 30th International Conference on Real-Time Networks and Systems (RTNS ’22) (2022, June)

Assessing traversal times is the main concern in the verification of embedded real-time networks. Schedulability analysis, as it provides firm guarantees, is the preferred technique for the designers of ... [more ▼]

Assessing traversal times is the main concern in the verification of embedded real-time networks. Schedulability analysis, as it provides firm guarantees, is the preferred technique for the designers of critical systems. There are however contexts where it is not economically or technically feasible to develop one, typically when the software and hardware components have not been designed with predictability in mind, e.g. as soon as TCP-based traffic is involved in network communication or when the hardware platform is too complex (e.g. heterogeneous System-on-Chips). In this paper, we study if it is possible to improve the ability of simulation to observe large traversal times, by running many short simulations with appropriately chosen simulation time and varying initial offsets of the stations on the network. The de-facto standard approach to assess maximal traversal times is to run a single long simulation with synchronized node start offsets and to use randomized node clock drifts inside an acceptable range. This approach is known to yield high traversal times but is not parallelizable. We propose an alternative approach consisting in splitting the simulation time over multiple shorter simulations with, optionally, randomized node start offsets. We evaluate the optimization potential of this simple approach on several realistic network configurations by comparing long simulations to aggregated short simulations, with and without synchronized node start offsets. We observe, considering all flows, that this allows a median improvement of up to 21.3% in terms of maximum traversal time observed, for the same simulation time budget. Additional randomization of the node start offsets showed further improvements of up to 4.8% in our experiments. Results from this line of work can be used to estimate the pessimism of schedulability analyses and verify systems for which no analysis is available. [less ▲]

Deep Learning to Predict the Feasibility of Priority-Based Ethernet Network Configurations

in ACM Transactions on Cyber-Physical Systems (2021), 5(4), 126

Machine learning has been recently applied in real-time systems to predict whether Ethernet network configurations are feasible in terms of meeting deadline constraints without executing conventional ... [more ▼]

Machine learning has been recently applied in real-time systems to predict whether Ethernet network configurations are feasible in terms of meeting deadline constraints without executing conventional schedulability analysis. However, the existing prediction techniques require domain expertise to choose the relevant input features and do not perform consistently when topologies or traffic patterns differ significantly from the ones in the training data. To overcome these problems, we propose a Graph Neural Network (GNN) prediction model that synthesizes relevant features directly from the raw data. This deep learning model possesses the ability to exploit relations among flows, links, and queues in switched Ethernet networks, and generalizes to unseen topologies and traffic patterns. We also explore the use of ensembles of GNNs and show that it enhances the robustness of the predictions. An evaluation on heterogeneous testing sets comprising realistic automotive networks, shows that ensembles of 32 GNN models features a prediction accuracy ranging from 79.3% to 90% for Ethernet networks using priorities as the Quality-of-Service mechanism. The use of ensemble models provides a speedup factor ranging from 77 to 1715 compared to schedulability analysis, which allows a far more extensive design space exploration. [less ▲]

QoS-Predictable SOA on TSN: Insights from a Case-Study

Scientific Conference (2021, February 09)

This work is about the design and configuration of service-oriented communication on top of Ethernet TSN. The first objective is to present takeaways from the design and implementation of the Renault E/E ... [more ▼]

This work is about the design and configuration of service-oriented communication on top of Ethernet TSN. The first objective is to present takeaways from the design and implementation of the Renault E/E Service-Oriented Architecture (SOA) called FACE. In particular, we discuss technological, design and configuration choices made for the SOA, such as how to segment messages (UDP with multiple events, TCP, SOME/IP TP), and the technical possibilities to shape the transmission of the packets on the Ethernet network. The second objective is to study how to ensure the Quality of Service (QoS) required by services. Indeed, services introduce specific challenges, be it only the sheer amount of traffic they generate and if there is a growing body of experiences in the use of TSN QoS mechanisms most of what has been learned so far is mostly about meeting the requirements of individual streams. Less is known for services that involve the transmission of several, possibly segmented, messages with more complex transmission patterns. We show on the FACE architecture how SOME/IP messages were mapped to TSN QoS mechanisms in a manual then automated manner so as to meet the individual requirements of the services in terms of timing, and the system’s requirements in terms of memory usage. [less ▲]

Towards Computer-Aided, Iterative TSN-and Ethernet-based E/E Architecture Design

Scientific Conference (2020, September 16)

In this presentation we would like to propose a novel approach towards studying and simulating candidate designs of next generation Ethernet architectures at established OEMs that intend to employ 100BASE ... [more ▼]

In this presentation we would like to propose a novel approach towards studying and simulating candidate designs of next generation Ethernet architectures at established OEMs that intend to employ 100BASE-T1, 1000BASE-T1 and, for increased flexibility and lower cost, 10BASE-T1S. Typical design goals of next generation architectures are future extensibility and cost optimization of the lowest-end. We propose to introduce guidance to an otherwise standard Monte-Carlo simulation by providing certain fixed points (e.g., mandated connections of ECUs to certain bridges, complete re-use of ECUs) and “hot spots” in the topology (e.g., ECUs with the highest variability pressure) that are known in advance from BMW’s experience with their vehicles in the field. Several important practical considerations must be integrated in the generation of candidate architectures: - Topological constraints: ECU proximity to sensors, daisy chain connections between ECUs to minimize cable length, number of switch ports in a certain ECU, etc. - Security and reliability requirements: segregation between mixed-criticality streams, proxy ECUs, and redundant paths. Our position statement explores the ability of algorithmic tools to synthesize Ethernet-based architectures based on a minimal fixed core TSN topology, design goals, design constraints, assumptions about next generation applications and data from past projects (capturing part of the OEM domain knowledge). [less ▲]

Early-stage Bottleneck Identification and Removal in TSN Networks

Scientific Conference (2020, February 12)

There has been a pivotal change in the design of E/E architectures, which is that OEM cannot assume anymore that the functions, and thus the communication requirements, are known in advance and fixed over ... [more ▼]

There has been a pivotal change in the design of E/E architectures, which is that OEM cannot assume anymore that the functions, and thus the communication requirements, are known in advance and fixed over time. It has become crucial for OEMs to be in the position to add further functions and services during the lifetime of the vehicle: OEMs need to design E/E architectures that are future-proof. We propose design space exploration algorithms, which, by answering a series of design questions and proposing solutions to the designers, help to improve an automotive E/E architecture in several dimensions: performance, costs, reliability and extensibility. We start by estimating the total "capacity" of a baseline architecture, then, by removing bottlenecks, we obtain an "enhanced capacity" architecture, which is then optimized in terms of costs and reliability. The architecture is then further optimized in terms of costs and reliability. The work is conducted on Volvo's prototype centralized and domain-based E/E architecture. [less ▲]

Early-stage topological and technological choices for TSN-based communication architectures

Scientific Conference (2019, September 24)

A main issue in the design of automotive communication architectures is that the most important design choices pertaining to the topology of the networks and the technologies to use (protocols, data rate ... [more ▼]

A main issue in the design of automotive communication architectures is that the most important design choices pertaining to the topology of the networks and the technologies to use (protocols, data rate, hardware) have to be made at a time when the communication requirements are not entirely known. Indeed, many functions only becomes available along the development cycle, and vehicle platforms have to support incremental evolutions of the embedded system that may not be fully foreseeable at the time design choices are made. The problem is becoming even more difficult and crucial with the introduction of dynamically evolving communication requirements requiring network re-configuration at run-time. We present how the use of synthetic data, that is data generated programmatically based on past vehicle projects and what can be foreseen for the current project, enables the designers to make such early stage choices based on quantified metrics. The proposals are applied to Groupe Renault's FACE service-oriented E/E architecture with the use of a software-implemented function we called “Topology Stress Test”. [less ▲]

Using Machine Learning to Speed Up the Design Space Exploration of Ethernet TSN networks

Report (2019)

In this work, we ask if Machine Learning (ML) can provide a viable alternative to conventional schedulability analysis to determine whether a real-time Ethernet network meets a set of timing constraints ... [more ▼]

In this work, we ask if Machine Learning (ML) can provide a viable alternative to conventional schedulability analysis to determine whether a real-time Ethernet network meets a set of timing constraints. Otherwise said, can an algorithm learn what makes it difficult for a system to be feasible and predict whether a configuration will be feasible without executing a schedulability analysis? In this study, we apply standard supervised and unsupervised ML techniques and compare them, in terms of their accuracy and running times, with precise and approximate schedulability analyses in Network-Calculus. We show that ML techniques are efficient at predicting the feasibility of realistic TSN networks and offer new trade-offs between accuracy and computation time especially interesting for design-space exploration algorithms. [less ▲]

Insights into the performance and configuration of TCP in Automotive Ethernet Networks

Scientific Conference (2018, October 10)

TCP has become an integral part of the Autosar communication stack located immediately below the Socket Adaptor layer. There are numerous use-cases for TCP in automotive networks. First TCP can be ... [more ▼]

TCP has become an integral part of the Autosar communication stack located immediately below the Socket Adaptor layer. There are numerous use-cases for TCP in automotive networks. First TCP can be leveraged on by many higher-level protocols be it for diagnostic (DoIp), calibration protocols (XCP), service-oriented communications (SomeIP), FTP and HTTP transfer, or more specialized tasks like protocols for the control of electric vehicle charging. In addition, communication through standard TCP sockets simplifies the development and re-use of applications requiring reliable transmissions within the vehicle or with external end-systems (car2x, cloud-based services).If previous works focused on throughput as performance criterion, we here also consider the communication latencies and memory usage in the end-systems and switches. The first question explored by simulation is to quantify the importance of TCP configuration choices in the latencies that can be achieved with TCP. In particular, it is shown that turning off Nagle's algorithm in transmission and delayed acknowledgment in reception as Autosar is very beneficial in terms of latencies but at the expense of throughput. We then extend the experiments to integrate the interactions between the Socket Adaptor and TCP layers. Finally, we study the performance of TCP streams as part of a complete automotive TSN case-study made up of control traffic, audio/video streams and best-effort traffic to evaluate the impact of higher priority traffic classes on TCP streams throughput and delays. The contribution of this work is twofold. First, we draw up a panorama of the TCP features and configuration options available in Autosar. Second, through experiments on models increasingly close to an implementation, we provide quantified insights in what we can expect from TCP in terms of latencies and throughput with the different configuration options available to us. [less ▲]

Towards the Systematic Analysis of Non-Functional Properties in Model-Based Engineering for Real-Time Embedded Systems

in Science of Computer Programming (2018), 156

The real-time scheduling theory provides analytical methods to assess the temporal predictability of embedded systems. Nevertheless, their use is limited in a Model-Based Systems Engineering approach. In ... [more ▼]

The real-time scheduling theory provides analytical methods to assess the temporal predictability of embedded systems. Nevertheless, their use is limited in a Model-Based Systems Engineering approach. In fact, the large number of applicability conditions makes the use of real-time scheduling analysis tedious and error-prone. Key issues are left to the engineers: when to apply a real-time scheduling analysis? What to do with the analysis results?} This article presents an approach to systematize and then automate the analysis of non-functional properties in Model-Based Systems Engineering. First, preconditions and postconditions define the applicability of an analysis. In addition, contracts specify the analysis interfaces, thereby enabling to reason about the analysis process. We present a proof-of-concept implementation of our approach using a combination of constraint languages (REAL for run-time analysis) and specification languages (Alloy for describing interfaces and reasoning about them). This approach is experimented on architectural models written with the Architecture Analysis and Design Language (AADL). [less ▲]