The impact of autonomous driving on software & hardware needs
Higher levels of automation require more hardware devices, software development and robust data processing
The automotive industry’s megatrends — i.e. electrification, automated driving, smart mobility and connectivity — are rapidly increasing the range and the complexity of functions required in modern vehicles. Therefore, carmakers and suppliers need to increasingly turn to software development to address the challenges.
Software systems embedded in modern connected & autonomous vehicles – i.e. driving assistance features, connectivity and infotainment product systems – are expected to grow rapidly in the forthcoming years. At the same time, consumers’ experience with cutting-edge technologies embedded in smartphones triggers expectations and demand for in-vehicle software innovative features. The move towards fully automated driving pushes for more dedicated software; as a large part of hardware-oriented systems becomes standardized and commoditized.
In addition, as connectivity functions are growing rapidly and vehicles become part of the internet of things (IoT) era, more software-oriented systems will be required to process, manage and distribute the vast amounts of generated data.
Intel, the tech giant that bought software supplier Mobileye, estimates that driverless vehicles will use roughly 4,000 gigabytes (GB) of data a day. Rapid and secure communication between the in-vehicle embedded electronic systems, between vehicles, and between vehicles and infrastructure are essential parts of the assisted and autonomous driving ecosystem.
ADAS and intelligent connectivity systems generate high demand for computing speed and data processing as vehicles integrate data from a growing number of in-vehicle ECUs.
Electronic Control Units (ECUs) are deployed on vehicles in order to enable digital control of functional electrical subsystems, such as steering and braking, powertrain control module, telematics, navigation, ADAS etc.
By integrating a microcomputer and a controlling embedded system, the ECU fuses raw data from multiple data sources including cameras, radars, ultrasonic sensors, as well as map data; a single ECU is dedicated to a separate function, thus, it handles only a small proportion of in-vehicle generated data.
The growing number of ECUs with sophisticated software escalates the vehicle system design complexity
The modern smart and connected vehicle requires more individual ECUs to support an increasing number of complex vehicular functions, higher integration between ECUs and in-vehicle sensors (cameras, lidar, radars, etc.), improved communication networks and transmission rates of the real-time information procedure to support the increasingly intelligent transportation systems.
This comes along with a growing number of ECUs, which is upwards of 100 on some vehicles today, executing around 150 million lines of code. The ECU’ implementation, along with the wiring, the plug-ins and the power supply requirements, leads to significant cost increases, higher in-vehicle SW complexity and constraints regarding vehicular functions’ optimization
The computing power required for a Lv.2 vehicle is 10 TOPS (tera operations per second), for a Lv.4 exceeds 100 TOPS, and it is estimated that a Lv.5 vehicle will require a 1,000 TOPS computing processing power.
The next-generation autonomous vehicle needs to integrate new, diverse technologies and complex logical operations; i.e. high-definition displays, advanced gateways, higher bandwidth; data fusion, full object detection, environmental modelling etc; and to leverage large amounts of generated data that require increased computing processing power.
In addition, the hardware architecture has to support advanced software functionality and upgradability; the latter is crucial to ensure that autonomous driving systems are not outdated and compile with modern safety and security standards.
From a large number of “one-function per ECU” to a more “centralized architecture” approach
In order to address complexity and optimization constraints, carmakers and automotive suppliers have moved toward more centralized E/E alternatives, such as domain control units (DCUs), where a DCU controls several sub-domain ECUs (each sub-ECU processing to realize vehicular functions) and integrates them into a single, more cost-efficient domain unit – e.g. Aptiv’s Active Safety Domain Controller integrated by BMW, Visteon’s Domain Controller SmartCore, launched by Mercedes Benz; zone-oriented control units (ZCUs), where the vehicle is divided into zones and each zone integrates into a zone-ECU (the sub-ECUs does not perform any processing to realize a vehicle function) – e.g. Bosch’s zone-oriented E/E architecture; and cross-domain centralized units (CDCUs), where the functions of more than one domain are consolidated onto single ECU (similar to the DCU’s architecture)- e.g. Bosch’s vehicle-centralized ECU.
All these solutions, the domain, cross or zone controllers, reduce the number of ECUs in the vehicle, enable sufficient SW-HW integration and increase the functionality of software modules. Centralized solutions are widely adopted by carmakers and suppliers and proved to deal well with the safety and security aspects of vehicle functionality.
Nevertheless, the ADAS and infotainment domain remain still a hot spot for centralized implementation as -not only- poses hard real-time requirements in terms of performance and computing, but also requires harmonization of hardware-developed systems to enhance the continuous upgradability of the burgeoning software applications.
Trends toward electrification and smart-connected technologies will result in significant growth in the automotive electronics and software market
According to a McKinsey report, the market for ADAS/AD sensors is expected to grow by 12 per cent p.a to reach $ 43 billion in 2030; mainly driven by the evolution in the LiDAR and the radar market. McKinsey predicts that the high market growth in the software and electronic market will be attributed to the increasing penetration of xEVs; i.e. the vehicles that combine higher SAE AV levels and an electric powertrain.
The production vehicle base of SAE level 3 cars with an electric powertrain is expected to grow significantly until 2030, enabling scale effects and better cost distribution regarding the cost of the electronic and software components.
The growing base of cars equipped with AD systems or electrified powertrains highly relying on Software and Electronics will drive market growth and open opportunities for companies.
McKinsey
Case study: Huawei software operating system in BAIC’s ARCFOX pure electric premium SUV
Huawei proposes a CC architecture based on computing and communication, comprising intelligent connectivity, robust vehicle control and smart driving; which is a cross-domain converged architecture scheme.
- Harmony OS smart cockpit connected system;
- Kirin 990 in-vehicle chip that can reach 800 TOPs computing power and supports 5G network. The chipset has an octa-core CPU comprising four large cores dubbed Taishan V20 Lite, an octa-core ARM G-76 GPU and a tri-core AI dedicated unit;
- Three 96-line automotive-grade semi-solid lidars, camera, domain controllers, and precise control algorithms; support Lv.2.5 autonomous driving in all scenarios of urban roads; NCA mode autopilot enables point-to-point automatic driving through the preinstalled map in the car in selected areas, including Beijing, Shanghai, Guangzhou, and Shenzhen.
Electric Vehicles and Embedded Software Systems
EVs provide a unique opportunity as a platform for software engineering systems; their functional procedures regarding wiring, heating, and energy management allow a more central computing architecture approach in comparison with ICVs.
Furthermore, EVs’ mechanical systems can be easily scaled down to create small versions of vehicle operational systems. With the EV as a software systems deployment platform, several aspects of vehicle-based software can be developed as unique engineering projects, with the ultimate end result being a “drivable” software system. In addition, electric vehicles allow developers to achieve greater architectural leapfrogging; following Tesla’s unique paradigm a number of Chinese and US firms are looking to leverage EVs’ straightforward architecture.
The Chinese “disruptors”: XPeng, LiXiang, Nio, BYD
Xpeng P7 integrates XPeng’s XPILOT 3.0; an autonomous driving system that uses a combination of radar, cameras, high-precision maps powered by Alibaba, localization systems and, most recently, lidar to detect and predict the road conditions and to accurately conduct automatic navigation assisted driving from A to B based on the navigation route set by the driver.
Li Auto’s Li ONE six-seat large premium SUV, is an extended-range electric vehicle (EREV) equipped with hybrid outlets that enable battery recharging through a petrol or a diesel generator. Li ONE is equipped with two Horizon’s latest “Jingcheng®3” dedicated AD chips that enable full-scene automated parking, visual perception, sensor fusion, etc.
Nio’s ES8 flagship SUV features NEDC range of up to 580 km, a NIO Pilot (advanced driver assistance system), and a NOMI in-vehicle AI system. NIO Pilot, empowered by Mobileye EyeQ4, provides over 20 driver assistance features and supports firmware-over-the-air updates. In addition, NIO adopts a battery-as-a-service (BaaS) model that allows users to swap batteries or have a charging truck directly to their homes.
Launched in 2020 BYD’s luxury electric sedan Han, quickly became the eighth most popular among the best-selling EVs in China. HAN eclectic sedan was the first BYD’s model equipped with DiLink3.0 system; an intelligent driving assistant system that integrates the ADAS module with DiTrainer; a big-data algorithm system that can automatically “observe” and “learn” driver’s behaviour, determine the driver’s type and, depending on the external driving environment, promptly remind the driver to turn on or to adjust the AD assistance functions.
How do carmakers respond?
The new-generation vehicles involve processes such as V2X communications, high-precision maps, high levels of automated driving and broader connectivity functions that require sophisticated algorithm calculations through high computing chips and deeply integrated SW and HW systems.
This procedure inevitably increases the SW/HW development costs; triggering carmakers to mitigate strategies to confront them. These strategies vary from adopting a cooperative approach, i.e. implementing partnerships with other OEMs and software-oriented tech providers; to developing differentiating parts of the SW architecture exclusively in-house.
New partnerships and cooperation models are emerging
In the first approach, new partnerships and cooperation models are emerging – primarily for software and electronics parts that have great innovation prosperities and are capable of providing greater product differentiation in comparison with hardware systems. Instead of vertical cooperation with dedicated suppliers, OEMs opt to partner tech providers in various development fields, such as high computing chips, cloud computing platforms, Lidar technologies, etc.; in order to create value and split development costs.
On the other hand, tech companies are seeking to cooperate with OEMs and component suppliers since, in most cases, they have neither the experience in building hardware nor the established sales market to offer an end-to-end product availability comparable to that provided to their tech consumers.
Examples in this direction include Ford’s cooperation with Lidar provider Velodyne, Volvo Cars partnership with Google, Mercedes partnership with Nvidia on SW-defined vehicles, SAIC and Luminar collaboration on Lidar and SW integration, tech giants Baidu and Huawei collaborations with multiple Chinese car makers etc. This approach favours cost reductions, but may lead to giving up power or relying exclusively on the provider’s roadmap; disrupting the harmonization between SW/HW development.
Built in-house software systems for in-car and vehicle-related services to develop differentiating software parts
In an effort to replicate Tesla’s successful model of developing in-house SW systems, traditional car makers like VW, opt to develop their one in-house SW platform. This approach promotes harmonized integration of future embedded hardware and software systems but needs a large number of vehicles in order to mitigate the increasing per/unit development cost.
VW has recently announced that it will develop its own standard software platform and will boost the in-house share of software development, in order to achieve a 60 per cent in-house software production by 2025.
According to the Group’s announcement, by 2025, all new Group vehicles will be equipped with VW’s operating system “vw.os” and will be connected to the Volkswagen Automotive Cloud. In order to confront production complexity issues, emerging especially from the SW-HW integration, VW plans to produce a smaller number of versions in which the individual configuration will no longer be set through the hardware when the vehicle is purchased; instead, desired functions will be available on-demand at any time using the in-vehicle digital ecosystem.
Toyota has recently established the Woven Planet Holdings Group, a company that focuses on the development of a more agile “software-first” in-vehicle ecosystem for future Toyota vehicles. The company has developed an Automated Mapping Platform (AWM), a connected crowdsourced software platform that supports the creation and distribution of HD maps. In addition, Woven currently integrates Toyota’s operating platform Arene OS; in order to enhance its ADAS, connectivity and cockpit capabilities.
Traditional Chinese carmakers, among them FAW, Geely and Changan are also forming teams in order to strengthen their software capabilities; whereas SAIC Motor launched its new Service-Oriented Architecture (SOA) software platform in Ap. 2021 and formed the SAIC Lingshu in 2020, a software development centre that will mainly focus on intelligent driving systems, software architecture, data processing, AI applications etc.
Read More
- The 10 most important Partnerships In Autonomous, Electric & Shared Mobility In 2021
- The 12 most important technology partnerships in Connected & Automated Driving in 2020
- Five New Revenue Pools for Automated Mobility
To learn more about competition and market developments in the next decade read our reports
