Summary:
This article mainly elaborates on the development process of ADCU from hardware design to software design, mainly including demand scenarios, key hardware circuits, circuit reliability, AUTOSAR architecture, CAN communication introduction, CAN communication software design, etc. Finally, a product-level intelligent driving domain controller is developed based on the above hardware and software technologies.
// Research Status of Intelligent Driving Domain Controller //
With the advent of the era of intelligent networking of automobiles, the electrical and electronic architecture of automobiles is also constantly upgrading with the times. At the same time, with the continuous improvement of the electronic level of automobiles, the electronic control unit (Electronic Control Unit, ECU), which is the predecessor of the domain controller, has gradually occupied the whole body of the automobile. Whether it is from the traditional engine control of the car, the anti-lock braking system, the electronic power steering system, the air conditioner, the entertainment audio-visual system, etc., all need ECU control. The number of body electronic control units is increasing, and some vehicles even have more than 100.
Bosch divides the automotive electronic and electrical architecture as shown in Figure 1. The increasing number of ECUs on the car makes the vehicle circuit too complicated. At the same time, with the advent of the era of intelligent driving, the number of on-board sensors is also increasing. After the one-to-one correspondence between the sensor and the ECU, the performance of the ECU will be greatly reduced. At the same time, the application of the ECU to intelligent driving will face Many technical problems, at this time domain controller (DCU, Domain Control Unit) just came into being. Its concept was first proposed by Tier 1 led by Bosch, Continental and Delphi, which divides the whole vehicle into Body and Convenience, Infotainment, Chassis and Safety, and Power System ( powertrain), and advanced driver assistance systems (ADAS) and other five major domains. Among them, the Automated Driving Control Unit (ADCU) is specially developed and designed for the vehicle intelligent driving system, which can receive multiple sensor data, such as camera, millimeter wave radar, lidar, and cloud data transmission V2X, etc., through VCU acquires vehicle dynamic data (vehicle speed, pedal signal, etc.). ADCU supports all input custom control strategies and execution decisions, and the output is used for driving status feedback and performing various intelligent driving functions on the vehicle.
Figure 1 Evolution of automotive electrical and electronic architecture
Kim H et al. used the Nvidia ARM platform with CPU+GPU architecture to implement DART, a DNN scheduling framework. In order to simplify the development and improve the portability between embedded applications, Soh J et al. used the hardware Zynq-7000 XC7Z045 platform to propose an unscented Kalman filter software-hardware co-design method. Lai YK et al. used the NVIDIA Jetson TX2 hardware platform to realize the target detection function. The hardware platform can meet the requirements of the target detection algorithm to select the target detection configuration and the given frame rate and detection accuracy. Lin Z et al. studied the deployment of field programmable gate array (FPGA) for traffic sign classification and detection, and verified that FPGA will achieve higher power efficiency than GPU.
Not only universities have conducted in-depth research on intelligent driving domain controllers, but enterprises also pay more attention to the development of intelligent driving domain controllers, as shown in Figure 2.
Figure 2 Industry overview of some intelligent driving domain controllers
Nvidia has always been in the leading position in chips. Since it entered the vehicle computing platform in 2015, it has provided basic computing power for intelligent driving. It has launched 4 products, namely Drive PX, Drive PX2, Drive AGX Xavier/Pegasus and Drive AGX/ Orin. At the same time, a new generation of SoC chip Orin will be launched in 2019. Compared with Xavier, it adopts the new Ampere architecture, and its computing power enters the top 200 intelligent driving domain controllers. NXP launched Blue-BOX in 2016 equipped with NXP S32V automotive vision processing and LS2088A embedded computer processor for modular management of sensors. Intel released Inter GO in 2017 equipped with two Mobilieye EQ5 camera chips, one for visual perception and one for information fusion. Inter GO uses FPGA as acceleration hardware and supports 5G communication at the same time.
As shown in Figure 3(a), the "DASy" developed by Bosch will become a centralized automotive architecture component, regardless of software and hardware, can be customized according to customer needs. Continental's ADC uses Renesas as its hardware architecture. For high configuration, using NVIDIA hardware architecture, ADC can support L2 to L5 intelligent driving. At the 2019 Shanghai Auto Show, ZF launched Pro AIRobotThink, which is applicable to L4-level intelligent driving, and it can configure the corresponding scenarios. As shown in Figure 3(b), the MDC platform launched by Huawei can be optimized through the integration of underlying software and hardware, and is now at the leading level in the industry in terms of time synchronization, minimum noise floor, low power consumption, and fast startup. As an open platform, MDC can provide standard and open API and SDK development kits. The Matrix 2 launched by Horizon supports 23-line segmentation. In complex environments, its monocular vision solution can output relatively stable perception results in extreme environments. The ADCU of Hongjing Zhijia optimizes the software and hardware system, so that the maximum power consumption of the whole machine does not exceed 100W. Its high efficiency and low consumption have very wide adaptability and can be applied to a variety of models.
Figure 3(a) Some international intelligent driving domain controllers
Figure 3(b) Some domestic intelligent driving domain controllers
// Intelligent driving domain controller design //
An Automated Domain Control Unit (ADCU) with good performance not only needs to adapt to long-term, high-frequency, and high-intensity environments, but also needs to choose a cost-effective microcontroller chip in order to adapt to more complex logic algorithms. . Therefore, for ADCU to adapt to different working conditions and environments, this article starts from the product-level domain controller design concept, and analyzes the ADCU design and development process, application scenario requirements, ADCU micro-control unit introduction, hardware circuit design, PCB reliability analysis and key software communication modules. Development and other aspects of design.
ADCU development process
The controller development process is mainly for the scientific decomposition of the entire controller development process. It includes product demand analysis, functional scheme design, reliability analysis, and also includes multiple links such as product hardware circuit design, product system debugging, and final product sealing and archiving. The sequence of different links requires continuous improvement and verification of the previous link. This paper mainly designs an intelligent driving domain controller. Its main function can not only read vehicle dynamic information and environment perception sensor information, but also perform data calculation on known information&#x