Our mobile future – Revolutionizing vehicle software architecture
Our relationship with our cars is changing fast and comprehensively. Emerging driver assistance functions, in-vehicle infotainment and mobility services, as well as driver-to-car and car-to-infrastructure connectivity affect the driving experience profoundly. On the road to fully autonomous vehicles, the driver’s role is rapidly changing from driver, to supervisor, to passenger.
When it comes to autonomous driving, the car must always be able to analyze the environment. Logically, it needs eyes via sensors to “see” what happens around it, to reliably take the passenger to their destination. Information on what the car “sees” is then analyzed by the brain – or AI, in this case – to make immediate decisions. And to send information quickly to the brain, we need a nervous system – or a network – which shares the information from one computer to others within the system.
To do this, let’s address 3 key engineering challenges:
1. High-performance hardware and network architecture Advanced driver assistance systems (ADAS) increase the software cross-dependencies in vehicles, requiring additional levels of computing performance and network communications far exceeding established vehicle electrical and electronic (E/E) systems. Massive sensor data processing (such as ultrasonic, RADAR, cameras and laser scanners) is necessary for object detection, object fusion (detecting people crossing the street or a red light) and trajectory calculation (where the car goes). 2. Electronic engineering and software development skillsets converging Developers need to utilize machine learning and create networks to connect the parts in the car, making this complex process require C++, Ethernet and security know-how to turn electrical software algorithms into high-performance codes. They also need integrated, high-performance Systems on Chip (SoC) to collect, analyze and manage large amounts of data. These include CPUs such as ARM Cortex A (v8) or Intel Atom (which can be found in Renesas R-Car, NPXi.MX or Intel Denverton Platforms). To handle these new requirements, electronic engineering and software development skillsets must converge. In addition, making mistakes during development delays time to market. The architecture has to change, causing a potential domino effect of issues to fix. To create and continuously update the software safely and with minimal errors, you need qualified experts. 3. Standardized procedures Since the skillsets of electronic engineering and software development are converging, the automotive supply chain is also becoming more complex. There are car manufacturers, vehicle device suppliers, chip producers like ARM and Intel, etc. To get from design to deployment, the supply chain needs standards. Without them, an effective supply chain is impossible – it would be too slow and expensive.
In response comes AUTOSAR (AUTomotive Open System ARchitecture), a worldwide development partnership of manufacturers, suppliers and others in the automotive industry that works to instill reliable safety regulations. With a collection of open, standardized software architecture and application interfaces, AUTOSAR aims to improve the performance of every vehicle and the experience of every user.
AUTOSAR – Standards enabling next-generation cars
Formed in 2003, AUTOSAR has adapted to industry changes over time. Because the car is becoming more futuristic with the digital cockpit, the communication between different parts – including applications, media and sensors – must be regulated. Just like how computers in an office communicate with one another, connected parts in the car must communicate reliably, safely and securely.
There are two different types of AUTOSAR regulations – Classic and Adaptive.
The Classic AUTOSAR platform applies to ‘old tech’. It focuses on hardware interfaces and electronic commands, such as simple electronic control units, real-time, safety and low-level software. Things like steering, braking, embedded actuator control and making sure sensors work fall under this category.
Alternatively, the Adaptive AUTOSAR platform answers to the emerging need for software and complex algorithms. It focuses on digital code and resulting actions (such as via machine learning), addressing the requirements for having Ethernet, a hypervisor operating system (OS), C++, cloud interaction and tight security. It’s basically the response to the need for more data processing through ultrasonic, RADAR, cameras, laser scanners and other methods.
Adaptive AUTOSAR also accelerates the hardware by installing advanced CPU models, more modern software and a much more flexible software distribution model (thanks to Linux, QNX, Android and others). Due to the rapid nature of the automotive market, the car must be adaptable to future connected devices, add-ons and updates. While not everyone has stepped into Adaptive AUTOSAR yet (as it is more of an IT realm), it’s essential to making a digital cockpit possible and stay at pace with innovative technological developments. We highlight an example in our previous blog, so be sure to give it a read.
Standards always work – rain or shine
With complex communications between various aspects of the car, there must be no room for error. For instance, there can be no time delays between visual, radar and laser sensors. Whether the car loses directional control, stops suddenly or speeds up, these actions put the user in danger and decrease their trust in the car (and the brand).
Take autonomous driving, for example. Combined with HMI solutions, we can incorporate virtual reality (VR) into the cockpit, allowing the user to explore the car’s capabilities interactively. But what does this mean for safety regulations?
The answers are: priorities, performance, security and flexibility.
Priorities Using AUTOSAR helps provide priorities. For instance, changing from automatic to manual mode sets priorities on interactions. First, the car asks the driver if they want to take control. If the driver doesn’t respond, the car slows down automatically, examines its surroundings and asks itself, “Is it safe to brake now?” This information flow is very complex, where the car must predict future actions by analyzing the environment and the human passenger (via motion or eye tracking sensors), resulting in an action that adheres to safety regulations.
Performance Future cars need high-performing networks and to be connected via Ethernet in order to work.
Security Since the car’s parts are connected to one another, a small misstep in security could affect the whole system. There’s one computer for sensors, one for mobility, one for apps, etc. With each computer linked to the others, hackers can “jump into” the network and control the wheel, the speed of the car, etc., since it’s all connected via the cloud. But Adaptive AUTOSAR standards help secure the car. It’s imperative to follow the standards in order to build trust in your brand and keep users safe.
Flexibility Future cars need flexible cross-domain application integration to receive updates over the air. Through the cloud, computers can share information with one another, handling processes swiftly and seamlessly.
New challenges are emerging, and Adaptive AUTOSAR is a key enabler in overcoming these challenges. The amount of ECU (electronic control units) in cars is increasing, making vehicle architecture shift to a structure where only a few HPCs (high performing computing), or ‘software on a chip’, control large groups of hardware (such as the cockpit, sensors, or the car’s lights). This in combination with environment-scanning sensors produces an enormous amount of data that has to be analyzed and responded to very quickly. For more information on how this data is used, check out our other blog.
An active member of AUTOSAR, Luxoft has ample experience in software architecture, embedded software and real-time data analysis. Our competences include C++, algorithms, Ethernet, Linux/Genivi, IT security, ARM, Intel, embedded OS for ASIL C/D, low-level drivers, safety, real-time architectures and hypervisors. We excel in using cross-domain mixed-criticality application integration using modern, distribution-based build chains that fit market needs in terms of flexibility on updateability (software updates over the air, or SOTA), such as Yocto. Our ongoing projects include automatic parking, object detection and our high-performance ADAS platform.
Luxoft has deep knowledge in skill areas that pure AUTOSAR Classic players lack. With experts in high-technology areas such as AI, IoT and Big Data, we have what it takes to integrate large amounts of complex software both quickly and securely – bringing futuristic interfaces to life.
To find out more information, please contact us by clicking here.
Kai Richter Senior Technical Director of Automotive, Luxoft
Is an electrical engineering with a PhD in computer engineering and the founder of Symtavision. After 10 years he sold the company to Luxoft where he currently works as engineering director in the automotive line of business. He creates dynamic software architectures, real-time operating systems, safety concepts, and complexity management. My passion is working together with people who like to drive things forward. On a personal level he enjoys wood working, cooking (and eating) and climbing volcanos.