IVN(In-vehicle network)[원문 발췌]

IVN(In-Vehicle Network)

IVN 기술의 차량 내 적용 사례 출처:

IVN 발전의 역사, 동향

Analysis of Frame Replication and Elimination for Reliability in Zonal-based IVN Architecture

• Traditional IVN Architecture: Then, with the introduction of OBD, a gateway, which is the first Ethernet application to vehicles, was placed to connect the ECUs and an external OBD connector for diagnostics

Example of traditional IVN Architecture

• In the future, more electronic devices and sensors are expected to be installed for new technologies such as autonomous driving, requiring higher bandwidth for low latency within the IVN. Accordingly, the current IVN architecture has evolved by utilizing Ethernet in the backbone network of the Domainbased IVN architecture to improve the performance of IVN

Example of domain-based IVN Architecture

• The ECUs in zonal-based IVN architecture are separated into zones, and connected to a zone controller that is responsible for separate control to the ECUs in each zone. And in the center of the architecture, there is a high-performance CCU not just a gateway. In Fig. 2.6., there is an example of zone-based IVN architecture. The CCU is responsible for the centralized control of a vehicle and manages the integration of the technologies applied to the vehicle. In addition, zonal-based IVN architecture provides more flexibility and determinism, making it easier to mount a ECU or apply new technologies.

Example of zonal-based IVN Architecture

Developed zonal-based IVN architecture

Review of Electrical and Electronic Architectures for Autonomous Vehicles: Topologies, Networking and Simulators

• the discussion on topologies of E/E architectures
  • However, as functions and computational demands continued to surge, the number of ECUs increased substantially, resulting in complex wiring harness systems with heightened weight and cost [8]. Hence, the production and integration of ECUs and wiring harnesses became one of the most expensive components in a vehicle, trailing only the power and chassis systems in terms of cost. To address these challenges, concepts like domain-oriented architectures, characterized by functional concentration, were introduced to reduce ECUs and enhance computing power.
  • In 2015, BOSCH presented a serial roadmap for the development of E/E architectures [9], a framework widely acknowledged and discussed in both academic and industrial circles. However, misconceptions and queries persist, particularly concerning the serial roadmap's perception that zone-oriented E/E architectures follow in the footsteps of domain-oriented E/E architectures.
  • Indeed, the serial roadmap appears challenging in satisfying the diverse requirements of Original Equipment Manufacturers (OEMs) or Tier1 suppliers due to their distinct development ecosystems and the varying demands of products and product series
• The discussion on networking technologies in the design of E/E architectures
  • In the traditional E/E architectures, E/E components are designed and developed in a distributed fashion, mainly using various bus technologies, such as CAN [10–14], CAN-FD [13–17], LIN [13, 14], FlexRay [13, 14, 18, 19] and MOST [13, 14], to connect in-vehicle ECUs for diferent E/E systems. These E/E/ architectures have been widely used in vehicles, such as BMW 7 series [20].
  • However, these bus technologies face limitations in bandwidth, characteristics, and applications, failing to meet the escalating demands of modern E/E architectures in autonomous vehicles.
  • The inadequacies stem from two principal aspects.
    • Firstly, none of the traditional bus technologies can adequately transmit high-defnition video streams requisite for autonomous driving or advanced infotainment.
      • For instance, a video stream with specifcations of 1920×1080 resolution, 24 bits depth, 60fps, and a 10-times compression ratio demands approximately 300Mbps bandwidth [21].
    • Secondly, many traditional bus technologies lack security features [22], real-time capabilities, and MACbased or IP-based scheduling capabilities, all of which are essential for modern vehicles [13, 14, 23].
  • In order to solve the above problems, new in-vehicle networking technologies have emerged.
    • For example, automotive Ethernet provides high-bandwidth communication capabilities
    • TSN technology provides good real-time guarantees, and SOA technology provides service-oriented capabilities
  • While previous literature has analyzed specifc vehicle networking technologies in, they are more focused on one networking technology [24–26]. There is a notable absence of comprehensive consideration for both vehicle system functions and network system design in existing literature.
• modeling and simulation technologies
  • these are critical in the design phase of E/E architecture, because they can signifcantly reduce development costs and time.
• Parallel Roadmap of E/E Architecture Topology Development
  • State of the Art of E/E Architecture Topologies in Academia and Industry
    • comprehensive examination of the current state of E/E architecture topologies in both academic and industrial domains
      • In academia, Stolz et al. [28] from BOSCH pioneered the integration of functions from multiple ECUs into DCUs, addressing the burgeoning complexity in E/E architecture topologies.
        • Navale et al. [9] from BOSCH frstly introduced a roadmap of the development of E/E architecture topologies with six main stages, including modular, integration, centralization, fusion, vehicle computer, and vehicle cloud computing stage
        • Brunner et al. [24] and Wang et al. [34] surveyed networking and communications technologies for autonomous driving and claimed that topologies of E/E architectures will evolve from a top-down fashion to a heterogeneous fashion.
        • Jiang et al. [1] discussed the improvements that need to be made in the E/E architecture topologies in order to adapt to the new trends in the automotive industry, including dedicated central gateways, domain masters, and zone-oriented structs. 
        • From the viewpoint of autonomous driving demand, Zhu et al. [36] analyzed the features, strengths, and weaknesses of point-to-point, vehicle bus-based, domain-based, zone-based E/E architecture topologies.
        • Dibaei et al. [37] proposed that the topologies of E/E architectures, as the underlying system technology of intelligent connected vehicles, will be developed towards a domain fashion into a centralized fashion.
        • Walrand et al. [38] designed a topology of zone-oriented architecture with a three-layer network, which includes a core, fast, and slow network.
        • Askaripoor et al. [40] reviewed the challenges and technologies that need to be solved in the process of confguring and integrating key applications into the vehicle central computer when the topologies of E/E architectures evolve from decentralized to centralized, such as challenges of software confguration and mapping for automotive systems, mapping techniques and optimization objectives for mapping tasks to multicore processors.
        • In a recent study, Deng et al. [41] studied the modeling and design methods from AVB to TSN, and proposed domain-based and zone-based TSN-based automotive E/E architectures
      • In industry, the evolution of E/E architecture topologies have received extensive attention from OEMs and Tier 1 suppliers [42].
        • Taking the Tesla model series as an example, Tesla Model S/X uses the traditional E/E architecture topology with a central gateway, and Model S has a wiring harness system with a length of 3000 m.
        • Model 3/Y adopts a zone-oriented three-domain E/E architecture, which greatly reduces its wiring harness to 1500 m and 100 m.
        • In order to meet the needs of OEMs, Tier1 suppliers, such as BOSCH [9, 49], Continental [24, 49], Vector [50], Aptiv [51], and Huawei [52], are actively promoting research and development related to E/E architecture topologies.
        • On the one hand, some Tier1 suppliers customize and develop platforms of E/E architecture for OEMs, such as Huawei’s CC E/E architecture, and Aptiv’s smart E/E architecture. On the other hand, some Tier1 suppliers are actively developing the core components required for the new architecture, such as DCUs.
        • Serial roadmap six Phase: modular, integration, centralization, domain fusion, vehicle computer with ZCUs, and vehicle cloud computing

A recognized serial roadmap for development of topologies of E/E architectures

• Misconceptions and Queries on Serial Roadmap
  • some misconceptions still exist in some publications and reports of OEMs and Tier1 suppliers R&D center, such as zone-oriented E/E architecture topologies are considered follow-up of domain-oriented E/E architecture topologies [23, 29–31].
  • Moreover, zone-oriented E/E architectures are considered equivalent to central-zonal E/E architectures, where a central vehicle computer is essential
  • In practical engineering contexts, a spectrum of variants has emerged, deviating from the prescribed serial trajectory in response to distinctive technical statuses and market objectives. unconventional topologies like that of the Tesla Model 3, featuring leftbody and right-body zone controllers along with a central control module, or Huawei's distinctive communication and computing architecture'
  • These variations highlight the substantial diferences in products, product series, technical statuses, and market objectives, challenging the universality of the serial roadmap.
  • Moreover, for zone-oriented E/E architectures, a central vehicle computer is not essential. For example, in the E/E architecture of Tesla Model 3 or the new E/E architecture of BMW [48], functions are distributed across ECUs rather than centralized within a singular vehicle computer. The serial roadmap, inherently sequential, fails to capture the nuanced evolution of zone-oriented E/E architectures, consequently fostering misconceptions and queries among OEMs and Tier1 suppliers.
  • Since the above misconceptions and queries exist in the previous serial roadmap, it is necessary to clarify these issues by considering a parallel roadmap for the development of E/E architectures.

Proposed parallel roadmap for development of topologies of E/E architectures

• Proposed Parallel Roadmap for Development of E/E Architecture Topologies
  • The proposed parallel roadmap encompasses three distinct stages: the traditional E/E architectures stage, the current parallel development roadmap of E/E architecture topologies stage, and the future concentrated and vehicle cloud E/E architectures stage.
  • domain-oriented route
    • Its primary stage is the domain concentration stage, characterized by the integration of functions from the central gateway and numerous ECUs within each functional domain into DCUs. This integration aims to reduce the number of ECUs within the functional domain, and these DCUs are linked with the central gateway through CAN or Ethernet [34]. 
    • Step1: The high-speed point-to-point Ethernet serves as the backbone network for interconnecting DCUs.
    • Step2: Subsequently, there is a continual integration among DCUs, exemplifed by merging powertrain DCU and chassis DCU into a motion-control DCU to enhance the management of vehicle dynamics and economy
    • Step3: Ultimately, all DCUs are amalgamated into a single vehicle computer, realizing the integration of computing power functions and the concentration of computing power

 

domain oriented route 3 stages

• zone-oriented route
  • Step1: In zone-centralization, ECUs are centralized in ZCUs, which share control functions.
  • Step2: In part-zone fusion, master-ZCUs handle most functions, while slave-ZCUs focus on network communication.
  • Step3: Full-zone fusion sees master-ZCUs forming a computing core for highperformance computing, communicating with surrounding functionally stripped ZCUs.

Zone-oriented route 4 stages

 

	Domain-oriented route	Zone-oriented route
Features	Functions of ECUs are gradually integrated into DCUs, cross-domain DCUs and vehicle computer	Components of sensors, actuators and ECUs are connected to ZCUs nearby, with functions partly and fully centralized
Ecological compatibility	Roadmap is relatively smooth, which satisfes the ecology of most OEMs and Tier1 suppliers	More suitable for new car manufacturers in automotive industry (e.g., Tesla), and new Tier1 suppliers1 (e.g., Huawei)
Scalability and interchangeability	Subject to more limitations, such as wiring harness, intra-domain communication load, etc	Support more functional variants, plug and play with strong replaceability
Functional safety and realtime property	Integrate into domain controller according to function, it is easier to design the ASIL functional safety level of the domain controller [55]. This task is mostly accomplished by Tier1 suppliers	Difcult for functional safety design of ZCU to be provided by traditional OEMs to Tier1 suppliers for separate design. ZCU has high requirements for functional safety and multi-core controllers
Backbone network requirements	Most of data transmission is realized in each functional domain with relatively low requirements on transmission rate and transmission characteristics of Ethernet backbone	All kinds of signals will be transmitted to target electronic unit through backbone Ethernet, which imposes requirements and characteristics on transmission rate of backbone network
Harness cost saving	Less reduction in length and weight of wire harness	Reduce wiring harness and improve wiring efciency and wiring automation

 

• Case Analysis of Domain‑Oriented and Zone‑Oriented E/E Architecture
  • In the domain-oriented E/E architecture model, the backbone comprises four DCUs: DCU1 (infotainment), DCU2 (body), DCU3 (powertrain and chassis), and DCU4 (ADAS), along with a switch. Each DCU establishes connections with the ECUs within its functional domain, facilitating message interactions for function calculation and control.
  • Conversely, in the zone-oriented E/E architecture model, the backbone features four ZCUs: ZCU1 (infotainment and body), ZCU2 (ADAS and body), ZCU3 (body), and ZCU4 (body, powertrain, and chassis), alongside four switches. Each ZCU connects to the nearest ECU. If the corresponding computing function is housed in the ZCU, the ECU's message is directly transmitted to the cell for computation. In cases where the computing function is not available, the ZCU utilizes only the routing function to transmit the message to another ZCU.
  • Hybrid E/E architectures, illustrated in Fig. 6, are deemed to exhibit more comprehensive performance compared to both domain-oriented and zone-oriented counterparts.
  • This hybrid approach combines the merits of zone-oriented architectures by leveraging ZCUs to efectively minimize the number of ECUs.This leads to enhancements in wiring harness organization, scalability, and interchangeability. Simultaneously, hybrid architectures integrate characteristics from domain-oriented E/E architectures, which adhere to the principle of centralizing the calculation of coupled functions.

 

Domain-oriented E/E Architecture

Zone-oriented E/E Architecture

Functions	Message parameters settings
Powertrain	16 messages with 20 ms period
Chassis	20 messages with 20 ms period
Body	16 messages with 50 ms period, 2 streams with 100 Mbps
Infotainment	5 messages with 100 Mbps (AVB)
ADAS	10 messages with 100 Mbps (AVB, ST), 8 messages with 1000 Mbps (AVB)

Message types	Average WCTT with domain-oriented E/E architecture	Average WCTT with zone-oriented E/E architecture
Body	16.833 ms/16	21.842 ms/16
Chassis	25.52 ms/20	12.510 ms/20
Powertrain	16.8 ms/18	17.4 ms/18
ST Ethernet	0.402 ms/9	1.184 ms/9
AVB Ethernet	0.692 ms/14	1.674 ms/14

Performance comparison between two architectures

Performance comparison of three types of topologies of E/E architectures

A development roadmap of topologies of hybrid E/E architectures

• Networking Technologies for E/E Architectures
  • While existing literature has predominantly concentrated on the communication layer of networking technologies in the automotive domain, this section takes a more holistic approach by considering their integration into the automotive composite system, which comprises the vehicle control model, the model of automotive E/E architectures, and more.
  • From this contemporary standpoint, the discussion unfolds to elucidate the driving roles of TSN, SDN, SOA, and converged networking technologies in shaping automotive E/E architectures.
  • Automotive Ethernet
    • Consequently, automotive Ethernet, renowned for its high bandwidth and QoS, has gradually found application in automobiles [13].
    • Unlike traditional industrial Ethernet technology, which typically employs LAN technology, automotive Ethernet stands out by utilizing Unshielded Twisted Pair (UTP) instead of two pairs of Shielded Twisted Pair (STP) to meet the complex Electromagnetic Compatibility (EMC) requirements within a vehicle [63].
    • Presently, automotive Ethernet supports full-duplex transmission rates ranging from 100 Mbps, 1 Gbps, multiple Gbps, to arbitrated transmission of 10 Mbps. This is standardized by IEEE 802.3bw-2015 [64], IEEE 802.3 bp-2016 [64], IEEE 802.3ch-2020 [27], and IEEE 802.3cg-2019 [27, 65].
    • Automotive Ethernet, with extended Time-Sensitive Networking (TSN) protocol support, facilitates flow synchronization, flow management, flow control, and flow integrity. This capability ensures low-latency, low-jitter, and safe communication essential for autonomous driving [27, 41, 66, 67].
    • Operating at the upper layer with protocols like TCP or UDP, automotive Ethernet can employ the SOME/IP protocol. This protocol, standardized in 2016 by AUTOSAR [44], is built on a scalable service-oriented network middleware.
    • The high bandwidth, low delay, and low jitter characteristics of automotive Ethernet make signifcant contributions to the paradigm shift in automotive E/E architectures [36].
      • Firstly, the high bandwidth enables the in-vehicle network to transmit a greater number of audio and video streams.
      • Secondly, the low delay and low jitter characteristics ensure Quality of Service (QoS) for various types of messages.
  • Automotive TSN Technology
    • The Time-Sensitive Networking (TSN) protocol family, evolving from the AVB protocol family in 2012, is set to empower automobiles with high-bandwidth and high-realtime communication capabilities, in tandem with automotive Ethernet [41].
    • Currently, the industry is inclined to adopt the Credit-Based Shaper (CBS) protocol within TSN for automotive systems, primarily due to its lower costs in design and embedded development compared to other TSN protocols [38, 76].
    • Noteworthy features expected to augment automotive systems encompass guaranteed Quality of Service (QoS) [77, 78], clock synchronization [79, 80], communication redundancy [81], functional safety [82], and information security [83], among others.
    • An inherent advantage of TSN lies in its ability to classify messages efectively, ofering distinct transmission characteristics for control messages, audio and video messages, or best-effort messages. These advantages play a pivotal role in ensuring the reliability and stability of vehicle control functions, thereby upholding vehicle functional safety. Moreover, TSN provides optimal streaming characteristics for audio and video communication requirements, such as those for cameras and displays, while contributing intelligent features to vehicles.
    • Considered for deployment in both domain-oriented and zone-oriented automotive E/E architectures, TSN technology facilitates real-time and reliable message transmission over Ethernet as the backbone. Messages are directed to Domain Control Units (DCUs) or Zone Control Units (ZCUs) for computation, possibly facilitated through gateway routing units [41].
  • Automotive SDN Technology
    
    • Software-Defined Networking (SDN) has emerged as a networking technology in recent years, introducing a paradigm that decouples the control plane from the data plane. This separation enables the dynamic confguration of the network plane through network ports, catering to fexible service requirements.
    • SDN technology offers fexible control over functions in DCUs, ZCUs, or switches, allowing for comprehensive management of network conditions.
      • For instance, during high-speed vehicle operation, consideration can be given to increasing the transmission frequency of relevant messages to achieve higher performance, reverting to a normal state during low-speed operation.
  • Automotive SOA Technology
    • In contrast to the conventional communication matrix reliant on fxed message transmission intervals, ServiceOriented Architecture (SOA) technology, based on the SOME/IP protocol, dynamically defines transmission services according to real-time requirements.
    • This approach proves efective in curtailing communication loads and ECU power consumption.
    • Acting as a conduit between the underlying hardware and upper-layer service applications, SOA facilitates fexible invocation of upper-layer services.
      • For instance, with SOA, communication with light ECUs can be activated only when the lights are in demand. In the context of autonomous driving functions, whether it's the DCU of ADAS in domain-oriented E/E architectures or the ZCU integrated with ADAS functions in zone-oriented E/E architectures, decisions about activating functions like ACC [87] or AEB [88] are made. This includes determining whether communication services are needed to invoke other control units
      • thereby contributing to reduced power consumption in embedded platforms
• Future Converged Networking Technology
  • TSN technology primarily ensures the real-time performance of communication message transmission, characterized by low delay and low jitter.
  • SDN technology introduces a controllable management layer into in-vehicle network architectures.
  • SOA technology enhances the fexibility of services in E/E architectures by incorporating service-based features.
• Modeling and Simulation Technology for E/E Architectures
  
  • In this section, some commonly used simulators are discussed for design and verifcation, mainly including some topology and networking simulators commonly used in the communication feld, such as OMNET + + , RTaw-Pesage and PREEvision, and topology and networking’s application simulators in the automotive areas, such as Matlab. Besides, some future comprehensive simulators are envisaged.
  • Topology and Networking Simulator
    • OMNET + + , an open-source discrete-event network simulator since 1997, is adept at simulating various industrial networks, including automotive in-vehicle TSN networks [56–58, 79]. It enables the construction of an Ethernet model for the new functional domain-oriented and zone-oriented E/E architectures. While OMNET + + was not originally developed for automotive networks, particularly in-vehicle networks, it lacks simulation models for CAN, LIN, FlexRay, and other in-vehicle networks, along with corresponding gateway models.
    • TCN Time Analysis [96], developed by Time Critical Networks in 2017, serves as a software tool facilitating the construction of architecture and network digital twins. This tool allows designers of networks and automotive electrical architectures to conduct architecture exploration and what-if analyses through simulation. The latest version provides basic support for CAN, LIN, FlexRay, Ethernet, and some TSN protocols. Notably, while recent literature [92] suggests the use of TCN Time Analysis for validating TSN and SDN techniques, it is not an open-source network simulation framework, limiting the implementation of new extensions with protocols or features
    • RTaW-Pesage [20, 97], introduced by INRIA in 2007, stands as a commercial vehicle network simulator with widespread industry usage. It has garnered adoption by prominent entities such as Mercedes-Benz, NIO, Bosch, Aptiv, and Huawei. This simulator excels in constructing automotive networks, including CAN, LIN, FlexRay, Ethernet, and others. Notably, it supports the simulation of cutting-edge Ethernet protocol stacks, encompassing TSN and SOME/IP. Distinguishing itself from OMNET+ +and TCN Time Analysis, RTaW-Pesage actively incorporates new in-vehicle network technical features.
    • PREEvision [50], developed by Aquintos in 2008, emerges as a commercial model-based development tool tailored for the design and assessment of E/E systems. Setting it apart from OMNET+ +simulation software, PREEvision has the capability to design the entire vehicle wiring harness, adheres to AUTOSAR and ISO26262 standards for system design, and enables evaluations based on criteria such as technology, cost, weight, and scalability.
  • Topology and Networking’s Application Simulator
    • Matlab, a numerical simulation software developed by MathWorks in 1984, serves as a convenient tool for establishing and verifying vehicle control models within the E/E architectures. However, its precision in vehicle modeling may be limited. In contrast, Carsim, a specialized vehicle dynamic modeling tool developed by MSC in 1996, enhances simulation accuracy. The commonly employed co-simulation of Matlab and Carsim [98] is integral in simulating vehicle systems, particularly for investigating control systems in the automotive domain, encompassing ACC [99, 100], AEB [101], and more.
    • The tripartite co-simulation involving Matlab, Carsim, and Truetime [102–104] facilitates the construction of vehicle control models under diverse E/E architecture topologies. This approach allows for the examination of various control algorithms across diferent topologies and networking characteristics.
    • For instance, a recent study [104] delves into the analysis of the impact of diferent control algorithms on ACC function through the aggregation loop under a domain-based E/E architecture. It's worth noting that this co-simulation, while robust, lacks support for accurate network models and the simulation of emerging networking technologies such as TSN, SDN, or SOA.
  • Future Comprehensive Simulator
    • To comprehensively analyze E/E architecture topologies, a synergy with the aforementioned simulator software is imperative.
    • Among the commonly used combinations, PREEvision and Matlab stand out for creating more comprehensive E/E architecture simulation models. This involved setting topology, hardware architecture, functional network, and requirement layers in PREEvision, with the design function features implemented in Matlab.
    • Neubauer et al. [107] proposed a framework leveraging both PREEvision and Matlab to autonomously synthesize hardware-centric Matlab models from multiple PREEvision E/E architecture layers.
    • RTaW-Pesage and Matlab can be efectively employed in co-simulation, with CPAL serving as a domain-specifc language developed by RealTime-at-Work and the University of Luxembourg. CPAL, detailed in Ref. [108], plays a pivotal role in interfacing with RTaW-Pegase for modeling high-level protocol layers within the E/E architecture. Simultaneously, it interfaces with Matlab for the development of vehicle control systems within the E/E architecture.
    • Given the intricate characteristics of topology, network, and application, the future demands more comprehensive simulators for realistic simulation analysis.

A development of simulators of automotive E/E architectures

• Challenges and Open Problems
  • Co‑design of Multi‑networking Mechanisms
    • However, their incorporation into the automotive feld necessitates careful consideration of the unique attributes of vehicles. This consideration goes beyond the typical concerns of low latency and low jitter in many cases within automotive E/E architectures, especially those involving sensor-controller-actuator structures like DYC and ACC, the paramount consideration becomes the end-to-end aggregation loop delay.
    • Automotive network control involves various task trigger modes, including timetrigger and event-trigger modes, and network control loops, encompassing sensor-controller, controller-actuator, and actuator-controller (feedback) loops. These diferent factors also need to be considered in the process of networking design.
  • Ultra‑High‑Speed Automotive Networks
    • In the context of future centralized E/E architectures, ensuring the precision of communication delay boundary calculations mandates a primary network utilizing automotive Ethernet with a transmission rate of up to 10Gbps [38].
    • Additionally, in-vehicle optical fiber communication technologies, characterized by ultra-high transmission rates, low EMI, and lightweight properties, are recognized as pivotal components in the blueprint of future centralized E/E architectures [109].
    • Whether it be in-vehicle Ethernet technology, in-vehicle optical fber communication technology, or other ultra-highspeed in-vehicle network technologies, it is unequivocal that they have not yet reached a mature application stage. Notably, the PHY chip of 10 Gbps in-vehicle Ethernet, for instance, has seen development only in recent years [38].
  • Safety Enhancement and Security Enhancement
    • As the topologies of E/E architectures undergo changes, the sensor-controller-actuator loop within the automotive E/E system is also altered. This modifcation in the message loop necessitates a reassessment and reinforcement of functional safety.
    • The continuous integration of new networking technologies within the new architecture provides the in-vehicle network with a more fexible and sophisticated scheduling mechanism. However, this advancement also introduces heightened security risks. For instance, the utilization of IP-based SOA technology, while offering advantages, exposes the system to potential IP-based security threats such as DoS or DDoS attacks.
    • Consequently, there is an ongoing imperative to fortify security measures in the emerging topologies of E/E architectures.
  • Standardization of In‑Vehicle Networks
    • Notable examples include the collaborative efforts of the 802.1 working group and the AVNU alliance in advancing the Time-Sensitive Networking (TSN) standard, the ongoing progress of the AUTOSAR standard, and the continuous evolution of standards such as ISO26262 and ISO13400.
  • Application to Embedded Platforms
    • Currently, various OEMs and Tier1 suppliers endorse multicore heterogeneous DCUs and ZCUs, along with in-vehicle Ethernet switches that support the in-vehicle Ethernet interface.
    • The design of DCUs and ZCUs necessitates a thorough analysis and allocation of functional requirements tailored to diferent vehicle models. The implementation of multi-core and heterogeneous designs in DCUs and ZCUs is contingent on the prior assignment of functions.
    • Particularly, in the realm of Service-Oriented Architecture (SOA) technology, on-board services must be comprehensively designed in alignment with functional requirements to ensure real-time, safety, and security.
    •  Concerning in-vehicle switches, positioned as general network devices, there is an expectation for enhanced versatility. These switches are anticipated to support protocols like TSN IEEE 802.1Qav, IEEE 802.1Qbv, and others, while also accommodating top-level Software-Defned Networking (SDN) or SOA scheduling designs.
    • The deployment of technologies like TSN and SOA on embedded platforms underscores the critical role of automotive software architecture and design. Diferent embedded platforms, such as Microcontrollers (MCUs) or System-onChips (SoCs), exhibit distinct bottom-layer operating systems like Linux or QNX, AUTOSAR architecture for Central Processing (CP) or Application Processing (AP), and middleware for SOME/IP or DDS. These diverse platforms necessitate unique software frameworks for efective design and performance optimization
  • Cost and Performance Optimization
    • Each automobile manufacturer harbors specific optimization goals and procedures for E/E architecture, often deemed as intellectual property.
    • In terms of cost optimization, this paper aligns with the four methods proposed in Ref. [8]:
    • 1) physical architecture and topological optimization
    • 2) optimization of conducting medium
    • 3) variant optimization
    • 4) redesign-to-cost approach.
    • However, there are few examples of publicly available research on the cost and performance optimization of the overall E/E architecture, because it often involves some sensitive business information.
• Conclusions
  • Firstly, diferent from the conventional linear roadmap, this paper introduces a novel parallel roadmap for the evolution of automotive E/E architectures. This divergence aims to dispel misunderstandings embedded in the serial roadmap. Two distinct trajectories, namely the domain-oriented route and the zone-oriented route, are proposed to cater to the diverse requirements of OEMs and Tier1 suppliers. Each route ofers unique advantages; domain-oriented architectures enhance real-time transmission within domains, while zone-oriented architectures present advantages in wiring simplicity, reduced ECUs, scalability, and interchangeability. The prospect of hybrid E/E architectures, combining the strengths of both, emerges as a promising avenue for future applications.
  • Secondly, the paper delves into the applications of TimeSensitive Networking (TSN), Software-Defned Networking (SDN), and Service-Oriented Architecture (SOA) within E/E architectures. Anticipations for future integrated networking technologies, combining multiple paradigms, are articulated. This forward-looking perspective refects the evolving landscape of networking requirements in the automotive sector.
  • Thirdly, a detailed exposition on topology and network simulators for E/E architectures is provided. The paper introduces simulators for both architecture topology and networking applications. While current simulators ofer valuable insights, the envisaged future calls for comprehensive simulators capable of delivering more realistic simulation analyses. This shift toward more sophisticated simulation tools aligns with the growing complexities of modern E/E architectures.
  • Finally, the paper concludes by addressing the challenges and open problems looming over the development of E/E architectures. These challenges span the realms of technology, integration, and simulation.
  • The development of future E/E architectures in autonomous vehicles requires the continuous joint eforts of both academia and industry areas.

Reference 사이트 및 논문

The Evolution of In-vehicle Network Architectures