ECU（Electronic Control Unit）是电子控制单元，是汽车专用微机控制器。汽车电子控制器的作用是接收来自传感器的信息，进行处理，输出相应的控制指令给到执行器执行。汽车ECU的核心在于微处理器，微处理器包括MCU、MPU、DSP 和逻辑IC 等。ECU领先企业包括博世、电装、大陆、Aptiv、伟世通等。

        随着车辆的电子化程度逐渐提高，ECU占领了整个汽车，从防抱死制动系统、四轮驱动系统、电控自动变速器、主动悬架系统、安全气囊系统， 逐渐延伸到了车身安全、网络、娱乐、传感控制系统等。 汽车 ECU 迅速增加，高端车型里的 ECU 平均达到 50-70 个，个别车型ECU数量超过100。

        随着车载传感器数量越来越多，传感器与 ECU一一对应使得车辆整体性能下降，线路复杂性也急剧增加，此时 DCU（域控制器） 和 MDC（多域控制器）等更强大的中心化架构逐步替代了分布式架构。

        域控制器（DCU，Domain Control Unit）的概念最早是由以博世、大陆为首的Tier1提出，它的出现是为了解决信息安全，以及ECU瓶颈的问题。域控制器因为有强大的硬件计算能力与丰富的软件接口支持，使得更多核心功能模块集中于域控制器内，系统功能集成度大大提高，这样对于功能的感知与执行的硬件要求降低。加之数据交互的接口标准化，会让这些零部件变成标准零件，从而降低这部分零部件开发/制造成本。也就是说，外围零件只关注本身基本功能，而中央域控制器关注系统级功能实现。

图例：伟世通将仪表ECU和车机ECU整合为座舱域控制器SmartCore

        而自动驾驶域控制器的要求更高，一般要具备多传感器融合、定位、路径规划、决策控制、V2X、高速通讯的能力，通常需要外接多种摄像头（单目、双目）、多个毫米波雷达、激光雷达、IMU等设备。

        由于要完成大量运算，域控制器一般都要配备一个运算力强大的核心处理器，能够提供对智能座舱和不同级别自动驾驶算力的支持，业内有NVIDIA、英飞凌、瑞萨、TI、NXP、Mobileye等多个方案。利用处理能力强大的多核 CPU/GPU 芯片相对集中的去控制每个域，以取代以前的分布式汽车电子电气架构（EEA）。

        上图是博世的 汽车电子电气架构进化图，包括Modular，Integration, Centralization, Fusion, Vehicle Computer, Vehicle Cloud Computing六个层次。其中域控制器应用在第三层（Centralization），多域控制器应用在第四层（Fusion）。

        目前新车E/E架构设计已大量采用域控制器。以奇点iS6为例，采用了基于域控制器加车载以太网的网络拓扑结构，将车内电子架构划分为5个域，分别为：智能驾驶域、智能信息域、车身域、底盘域、动力域。同时奇点汽车采用集成式设计，也就是将所有传感器数据集中到智能驾驶域控制器。由智能驾驶域控制器做数据处理和决策规划，从而实现各种ADAS智能驾驶功能，包括自适应巡航、车道保持、自动泊车等。这意味着大多数ADAS/AD功能将由整车厂自己开发完成。

        根据技术专家“冷酷的冬瓜”的研究：纵观特斯拉三代车型，Model S、Model X 再到 Model 3 的演变，实质是功能的重分配，不断把功能从供应商手中拿回来自行开发的过程。而Model S的E/E架构设计，早已一步跨到了第五层 (Vehicle Computer层级) 。

        随着汽车E/E架构的演变进化，主机厂和汽车电子供应商的供应关系正发生深刻变革。由于汽车电子硬件走向集中化的趋势，汽车电子供应商数量将减少，同时域控制器供应商将更加重要。

        以座舱域控制器为例，一般会集成仪表和车机，未来则会逐步整合空调控制、HUD、后视镜、手势识别、DMS，甚至包括T-BOX和OBU。

        一辆自动驾驶汽车每小时产生的数据量将达4TB，传感器融合和3D定位等高级功能将在自动驾驶域控制器上运行。

        和域控制器紧密连接的中央网关，负责发送与接收安全关键数据，将始终直接且仅连接到整车企业的后台。OTA则实现域控制器功能的不断更新，有助于开发新功能、确保网络安全，并使车企得以更快部署功能与软件。

        域控制器厂商和车厂的开发合作将更加紧密。

        德赛西威认为：在自动驾驶域控制器领域，预计未来Tier1与整车厂之间将采取两种合作方式。

        其一，Tier1负责中间层以及硬件生产，整车厂负责自动驾驶软件部分。Tier1的优势在于以合理的成本将产品生产出来并且加速产品落地，因此整车厂和Tier1进行合作生产方式是必然，前者负责自动驾驶软件部分，后者负责硬件生产、中间层以及芯片方案整合。
其二，Tier1自己与芯片商合作，做方案整合后研发中央域控制器并向整车厂销售，例如大陆ADCU、采埃孚ProAI、麦格纳MAX4。

        从下面两个图表，可以看到无论是座舱域控制器，还是自动驾驶域控制器，都有控制器供应商和主机厂、域控制器供应商和芯片厂商紧密合作开发的趋势。

        根据佐思产研的预测，2025年全球汽车域控制器（座舱+自动驾驶）出货量将超过1400万套，2019-2025年均增长50.7%。

Electronic control unit (ECU) serves as an automotive computer controller. Automotive electronic controller is used to receive and process signals from sensors and export control commands to the actuator to execute. Microprocessors, the core of an automotive ECU, embrace micro control unit (MCU), microprocessor unit (MPU), digital signal processor (DSP) and logic integrated circuits (IC). The global ECU leaders are Bosch, Denso, Continental, Aptiv, Visteon, among others.

As vehicle trend to use more electronics, ECU is making its way into all auto parts from anti-lock braking system, four-wheel drive system, electronically controlled automatic transmission, active suspension system and airbag system to body safety, network, entertainment and sensing and control systems. Vehicles’ consumption of ECU then booms: high-class models use 50-70 ECUs on average, and some even carries more than 100 units.

When the one-to-one correspondence between the growing number of sensors and ECUs gives rise to underperforming vehicles and far more complex circuits, more powerful centralized architectures like domain control unit (DCU) and multi-domain controller (MDC) come as an alternative to the distributed ones. 

The concept of domain control unit (DCU) was initiated by tier-1 suppliers like Bosch and Continental as a solution to information security and ECU development bottlenecks. DCU can make systems much more integrated for its powerful hardware computing capacity and availability of sundry software interfaces enable integration of more core functional modules, which means lower requirements on function perception and execution hardware. Moreover, standardized interfaces for data interaction help these components turn into standard ones, thus reducing the spending on research and development or manufacture. In other words, unlike peripheral parts just playing their own roles, a central domain control unit looks at the whole system.  

Figure 1: Visteon integrates instrument ECU and head unit ECU into SmartCore cockpit domain controller

Autonomous vehicle requires domain controllers not only to be integrated with versatile capabilities such as multi-sensor fusion, localization, path planning, decision making and control, V2X and high speed communication, but to have interfaces for cameras (mono/stereo), multiple radars, LiDAR, IMU, etc.

To complete number crunching, a domain control unit often needs a built-in core processor with strong computing power for smart cockpit and autonomous driving at all levels. Solution providers include NVIDIA, Infineon, Renesas, TI, NXP and Mobileye. The scheme that powerful multi-core CPU/GPU chips are used to control every domain in a centralized way can replace former distributed automotive electric/electronic architectures (EEA).

Figure 2: evolution of Bosch E/E architecture. It has six layers, i.e., Modular，Integration, Centralization, Fusion, Vehicle Computer and Vehicle Cloud Computing. DCU is applied to the third layer (Centralization), and MDC the fourth (Fusion).

 In current stage, most new vehicles adopt DCU-based E/E architectures. In Singulato iS6’s case, a DCU + automotive Ethernet based network topology is used to divide E/E architecture into 5 domains: intelligent driving, smart cockpit, body, chassis and power; an integrated design allows fusion of all sensor data into the intelligent driving domain controller which is in charge of data processing and decision making to implement ADAS functions such as adaptive cruise control, lane keeping and automatic parking. All imply that automakers need to develop their own ADAS/AD systems. 

The study by “Cool Wax Gourd”, a technical expert’s Twitter-like Sina Weibo account, shows that: the evolution of three generations of Tesla models from Model S to Model X to Model 3, is actually a process of functional redistribution, namely, developing capabilities based on those from suppliers; Model S E/E architecture has been a fifth-layer one (Vehicle Computer) at the start.

As automotive E/E architectures evolve, there is a big shift in relationship between OEMs and automotive electronics suppliers, too. The trend for integrated automotive electronic hardware leads to the smaller number of electronics suppliers and the more important role of DCU vendors.

Being generally integrated with instrument clusters and head unit, a cockpit domain controller for instance, will be fused with air conditioner control, HUD, rearview mirror, gesture recognition, DMS and even T-BOX and OBU in future.

An autonomous vehicle that generates 4TB data an hour, needs a domain control unit to have some advanced competencies such as multi-sensor fusion and 3D localization.

Central gateway closely tied with domain controllers, takes charge of sending and receiving key security data, and is directly and only connected to the backstage of automakers. Through OTA updates to domain controllers, carmakers can develop new capabilities and ensure network security for faster deployment of functions and software.

DCU vendors and automakers will deepen their partnerships in research and development.

Desay SV argues that: tier-1 suppliers and OEMs will collaborate in the following two ways in the area of autonomous driving domain controller:

First, tier-1 suppliers are devoted to making middleware and hardware, and OEMs develop autonomous driving software. As tier-1 suppliers enjoy edges in producing products at reasonable cost and accelerating commercialization, automakers are bound to partner with them: OEMs assume software design while tier-1 suppliers take on production of hardware and integration of middleware and chip solutions.
Second, tier-1 suppliers choose to work with chip vendors in solution design and research and development of central domain controllers, and then sell their products to OEMs. Examples include Continental ADCU, ZF ProAI and Magna MAX4.

It can be seen from the two tables below that there is a tendency towards cooperation between controller vendors and OEMs, domain controller suppliers and chip vendors, in both cockpit and autonomous driving domain controllers.

Typical Autonomous Driving DCU Vendors and Their Customers and Partners

DCU, as a kind of OEM automotive electronics, usually takes over two years from design to mass production and launch. Most of the above suppliers are still researching and developing DCU. Aptiv and Visteon are far ahead of peers and have mass-produced DCU.

The global automotive DCU (cockpit + autonomous driving) shipments will exceed 14 million sets in 2025, with the average annual growth rate of 50.7% between 2019 and 2025, according to shujubang.com.

Throughout the DCU industry, Chinese companies have emerged strikingly in the past two years, such as Desay SV, Baidu, Neusoft, HiGO Automotive, COOKOO, In-driving, iMotion, etc., all of which now takes emerging and non-first-tier traditional automakers as their key clients.