Software expertise and experience are becoming crucial core competency that is in short supply in the auto industry. The growing amount of software functionality that are embedded in every new car model will require increased competency by OEMs, Tier 1 and software suppliers. Software platforms are a required strategy to be able to keep up with the software explosion that are coming from customer’s demand for connected cars and apps-centric infotainment systems and the many ADAS functions.
In this article, I will complete my summary on challenges and options facing the automotive industry, by exploring automotive software platforms, market and technology trends. This is a third in the series of columns I posted last month, focused on software platforms (see below).
First, let me sum up in the following table the status of key software platforms used in the automotive industry.
The first seven entries in the table were already described in the previous column. The remaining software platforms are summarized below in this table, and they are marked in red in the first column.
Note: AUTOSAR=AUTomotive Open System ARchitecture; AGL=Automotive Grade Linux; C-S=Cybersecurity; ECU=Electronic Control Unit; NCAP=New Car Assessment Program; SoC=System on Chip; SW=Software; TSP-Telematics Service Provider; UN=United Nations;
OTA Software Platforms Over-the-Air (OTA) software update technology was developed in the high-tech industry, but OTA products are becoming integral to the automotive software business. This makes it more important for the auto industry to manage or control the key OTA software platforms—with the Tier 1 suppliers taking the lead.
Harman is the leading supplier of OTA software update solutions due to acquisition of Red Bend and Symphony-Teleca in 2015. Aptiv has also entered the automotive OTA segment with an acquisition of Movimento in 2017. Several other Tier 1 suppliers are also providing OTA capabilities via acquisition or cooperation with high-tech OTA suppliers. GM, Ford, BMW and Tesla are in this category with likely additional OEMs doing so in the future. Tesla software updates are based on Harman’s OTA platform. Tesla leads in OTA and are currently doing over 10 major software updates per year for its ECUs—excluding many additional updates for its AutoPilot system.
Some auto OEMs are developing their own OTA platforms, usually by combining their own effort and available technology from established OTA platforms. Harman’s Redbend OTA software platform is considered the leader in automotive remote software updates.
New OTA technology is emerging that position OTA as software life-cycle management with focus on the create and use phases. This means the OTA software functionality is integrated with the software development platforms and as part of their maintenance activities. Aurora Labs is an example of an OTA company following this approach. The strategy is to position the OTA platform as software-prognostics tool to help identify and fix software bugs. This trend also shifts OTA platforms into a service business.
Cybersecurity Platforms Cybersecurity software need to protect multiple systems in the car. The connected gateway is prevalent in connected cars and needs a cybersecurity software client and usually built-in cybersecurity hardware too. The gateway may also have cybersecurity software to protect communication via the car’s bus network. The most important ECUs also need their own built-in cybersecurity software clients. There are many cybersecurity companies that focus on the auto industry with many being recent startups. Examples are Argus (acquired by Continental), GuardKnox, Karamba, Iredeto, Regulus Cyber, SafeRide Technologies and Trillium Secure.
Argus is considered the automotive cybersecurity leader. Regulus Cyber includes protection from GPS spoofing, which is a growing problem. SafeRide includes protection for automotive Ethernet networks.
Just as important as cybersecurity software clients is a cloud-based SaaS platform for cybersecurity, usually called a cybersecurity operation center (SOC). The cloud-based SaaS cybersecurity platform can be used without in-car cybersecurity software client for vehicle fleets. Upstream Security is one of the leading companies with such service in the auto industry. Some of the companies offering cybersecurity software clients also offer SaaS platforms that interact with their in-car software clients and/or cybersecurity hardware.
Cockpit Domain ECU Platform A cockpit domain ECU integrates the cockpit display functions and infotainment functions into a single system. The cockpit domain ECU may include multiple displays such cockpit display, center-stack display, rear view mirror display and head-up display. The infotainment system includes multiple systems such as audio components, a variety of user interfaces (knobs, touch, speech) and smartphone app integration. Telematics systems are likely to be included in high-end versions.
This integration requires more powerful software platforms and is built on high-performance SoC hardware platforms. The integration advantage comes from eliminating multiple ECUs and subsystems, which creates hardware cost savings, fewer parts and suppliers, weight and space savings.
The cockpit domain ECU software platforms are primarily based on the infotainment software platforms such as the operating systems, OTA, cybersecurity, Virtual Personal Assistant (VPA), smartphone apps integration and others. The OS by definition has to serve safety critical systems, which requires ISO 26262 certification. If Linux versions are used, it will require a hypervisor plus another certified OS.
The cockpit domain ECU is in the early growth phase since volume production started in 2017. The potential for software platform growth and SoC chips are significant and could reach tens of millions of units after 2025.
Visteon has been the early leader in supplying cockpit domain ECUs with Aptiv as another leader as they entered the market first. Many of the other leading Tier 1 suppliers such as Bosch, Continental, Harman, Marelli and Panasonic, are also ready with products and volume production.
ADAS Software Platforms Advanced Driver Assist Systems (ADAS) include many functions that help the driver. ADAS include four levels of SAE’s six autonomous driving levels—Level 0, Level 1, Level 2 and Level 3. L0 has no automation and consist of warning functions such as park assist, blind spots, lane departure warning (LDW), collision warning and driver monitoring. L0 functions are available on most autos sold in many countries.
L1 provide driver assist functions with limited control of the car. The three main functions are adaptive cruise control (ACC), lane centering and semi-automated parking. L2 provide partial automation, but the driver must always monitor the driving. L2 examples are traffic jam assist and limited autopilot. L3 primarily consists of advanced autopilots for specific environments such as highways, low-speed urban areas and self-parking. The driver can disengage from the driving functions but must be prepared to re-engage within a few seconds when requested by the L3 system.
All ADAS functions are defined by software that are receiving data from camera, radar and ultrasonic sensors. Low-cost lidars are emerging for L2 and L3 functions.
Regulation is an important factor in ADAS growth as UN has approved extensive safety standards, which are being implemented by many region-specific NCAP organizations. European NCAP are on an aggressive schedule.
ADAS domain ECUs are emerging with L0-L1 ADAS integration appearing first. The main focus is on ADAS domain ECUs for L2-L3 functions which have multiple SoCs and extensive software platforms.
If you need detailed information on ADAS deployment, IHS Markit has historical and forecast data by OEM brand and models for all ADAS functions including sensors, apps and suppliers for most vehicles.
Autonomous Vehicle software Platforms The autonomous vehicle software platforms will have many functions and will be quite complex. The virtual driver software will probably be the most complex platform with a large code space. Many companies are developing AV software platforms. To get an overview of the many participants the, linked article has a picture of key companies and their connections. This picture is from early May 2020 and already several changes have taken place. An update will be done later this fall.
The sensor fusion software is the other major AV platform that will be closely connected to the virtual driver software. This platform includes considerable AI software technology and there are lots of companies developing, testing and improving sensor fusion functionality or parts of the platform or the underlying algorithms. Most of the virtual driver software companies are also developing sensor fusion software platforms.
An AV tele-operation software platform is also expected to be important. Tele-operation is required in the California when AVs have no safety drivers and other regions are following this strategy. There are multiple tele-operation startup companies including Phantom Auto, Ottopia, Designated Driver and DriveU.
AV software will need an advanced operating system to manage all the software and extensive hardware portfolio including large amounts of sensors. QNX looks well positioned as an AV OS and it has selected by Nvidia and others.
It is important to remember that most OEMs will want to have their own version of an AV software platform. This has been standard operating procedure in the auto industry and is likely to continue with AV software platforms. One remaining question is whether and/or how much Tier 1s will be involved in an OEMs tailored AV software platform.
The AV software platform segment has so many elements and complexity and is evolving rapidly. As more information becomes available, a separate column on this topic is desirable in the future.
Automotive Software Cost Perspectives Cost estimation of automotive software is often an art form and it is not any easier for software in other industries. Let us first look at the big picture—what would it cost to develop all new software for a vehicle? I will assume that the total amount of software is 100M lines of code. All I need is the average cost to develop one line of code. There is lots of data for a variety of software segments. I think the most useful data for automotive software is from Phil Koopman in an internet post in October 2010.
The internet post said that embedded software cost per line of code is $15 to $40. At $40 you get robust, well designed code suitable for industry applications, which applies to the auto industry. Since this data is nearly a decade old, the cost has probably increased, but I will use $40 figure. This means that to develop 100M lines of code for auto grade systems would cost $4 billion. Hence it is clearly desirable to re-use and update existing automotive software platforms where possible.
The next software cost perspective is about software royalties per software client included in yearly car sales. The software client royalties are usually a few dollars or less per unit. Software royalty for an operating system such as QNX will range from less than $10 for a telematics system to over $15 for an infotainment system. These costs exclude software maintenance and updates, which are usually negotiated separately as part of the software development project.
The cost of automotive software development varies by ECU segment. Infotainment software developments are common and is needed for most new models. A high-end infotainment system for a new model would cost around $20M for the software and a similar amount for the hardware. The typical development time would be 2-3 years including all testing and verification.
What about AV software cost? We don’t know much yet as there are numerous uncertainties. However, the VC investments in AV software has already surpassed $20B, but not all of it has been spent yet.
AVs will have multiple software platforms that will have a royalty when sold and SaaS service fees for the rest of their life. The AV software royalty fees will be much higher than current software royalties—probably in the range of $150 to $300 per AV. However, the real payoff for AV software will be the monthly SaaS fees that will be received from all AVs on the road at any time. The AV SaaS fees on a yearly basis could be similar to the initial royalty fees. The AV SaaS fees could be scaled to miles driven to get lower fees for personal AVs.
Software summary The OEMs that can leverage and combine third party software platforms and expertise with their own software platforms will be the long-term winners in the rapidly emerging software defined cars, given lots of innovations are coming from high-tech software firms.
OEMs want more software control as it defines most of the cars’ functions and how the users interact with their cars. But the OEMs need to rely on many modular software platforms and innovations that the software suppliers provide.
The automotive software business model is shifting towards services and SaaS-based relations. This is good for software suppliers as their revenues are proportional to autos in-use, which can be up 10X larger than yearly sales. But the OEMs will require cost savings in the overall software expenditures through more reliable software. OEMs will also expect upgradable software functions that will generate more revenue through the lifetime of the software—as Tesla is already doing.
Egil has over 35 years’ experience in the high-tech and automotive industries. Most recently he was director of research at the automotive technology group of IHS Markit. His latest research was focused on autonomous vehicles and mobility-as-a-service. He was co-founder of Telematics Research Group, which was acquired by iSuppli (IHS acquired iSuppli in 2010); before that he co-founded Future Computing and Computer Industry Almanac. Previously, Dr. Juliussen was with Texas Instruments where he was a strategic and product planner for microprocessors and PCs. He is the author of over 700 papers, reports and conference presentations. He received B.S., M.S., and Ph.D. degrees in electrical engineering from Purdue University, and is a member of SAE and IEEE.