Safety, reliability and robustness are the foundation of an autonomous vehicle development platform

Building safe and reliable advanced driver assistance systems (ADAS) and autonomous driving systems is complex, and no manufacturer can develop all the necessary technologies alone.

Instead, there is a broad ecosystem of partners, each working on a different aspect of the required technology.

There is no doubt that the integration of these subsystems is critical to the success of the autonomous vehicle market. With the right development and testing platform, developers can iteratively design, build, and test their ideas, moving them from concept to a validated, production-ready design faster and more reliably.

A company that develops ADAS and autonomous driving development tools recently turned to Vecow, a leader in designing and producing industrial-grade computers for applications such as automotive and industrial AI, for help in creating hardware for its development and testing platform.

First, the platform needed to be designed for OEMs and Tier 1 suppliers. It also needed to provide integrated software and hardware for developing and validating system functionality and performance. The platform also needed to be able to collect, synchronize, and analyze data from a variety of cameras and other sensors.

With so many different cameras and sensors to be used, the platform will be critical, and it will also be important that the system can support a full range of interfaces. This ranges from serial ports to CAN, CAN FD, FlexRay, LIN, Ethernet (i.e., for LiDAR), USB, and more. Flexible hardware with enough I/O to provide a sufficient number of each interface will also be required. As the list of available sensors and cameras continues to grow, the required hardware interfaces and software drivers must be easy to integrate to quickly and easily support each new device.

One problem OEMs encountered in their initial efforts was the difficulty in finding a single industrial computing platform that would meet the needs of most of their automotive customers. Each automaker had to meet different specifications for different applications and target markets. Having to build a specific computing platform for each customer would significantly delay a company’s ability to deliver a system. It would also make it more difficult to support every development platform in the field. Creating custom systems would drive up their prices and limit their ability to compete.

Manufacturers also don’t want to use custom platforms. When integrating systems together, it can be difficult to determine the requirements of each unique system, especially when each vehicle application is different. With a custom approach, manufacturers must determine which appropriate computing systems must be implemented on a case-by-case basis. This can lead to development delays as engineering teams must evaluate available technologies, define custom systems, and then spend time integrating the custom systems before they can be tested and evaluated. The further complexity of custom designs can lead to frustration, especially as ADAS and autonomous driving technologies are evolving at an incredible pace, resulting in constantly changing applications and market requirements.

To address this problem, the platform company turned to Vecow to help define and build an industrial computing platform that would meet the customers’ varying hardware, software, power, and budget requirements.

ADAS and Autonomous Driving System Requirements

The complexity of ADAS and autonomous driving functions means that the platform must provide powerful computing and processing resources. Further complicating the processing is that the vehicle needs to process a large amount of data in real time to meet the strict response requirements of the vehicle at high speeds.

To meet the interface requirements, the hardware needs to have rich and flexible I/O. Specifically, the platform needs multiple USB 2.0 and 3.0 ports for cameras, multiple PCIe slots for expansion cards, and multiple Mini PCIe slots for connecting other hardware.

A cross-platform solution was also a must. While the system is primarily based on Windows and Linux operating systems (OS), it needs to support all in-house and third-party software that manufacturers use across different platforms, applications, and OSes.

Car power

Of particular concern is in-vehicle power consumption and power protection. Development and test platforms need to run in vehicles to collect real-world data. These systems also rely on GPUs, which typically have high power consumption. One challenge facing automotive applications is that higher-power subsystems can affect safety. Specifically, if ADAS and autonomous driving platforms consume too much power, it may cause the system to crash or shut down, putting the driver in danger. Therefore, the platform needs to be able to operate reliably and robustly using the available power of the vehicle.

The system architecture also needs to consider physical robustness. Given the wide range of environmental conditions that the vehicle may encounter during road testing and actual use, the system must support a wide range of operating temperatures. The hardware must also be able to withstand the levels of vibration and shock that may occur during use, such as when the emergency brake is activated.

Optimize a consistent and flexible platform

Vecow has worked with partner companies to define three platforms of hardware and software solutions for its customers, called Autoware platforms. The entry-level platform is based on Vecow's SPC-4020A embedded PC, based on an Intel Atom CPU. It is designed for applications that require the collection of relatively small amounts of data and features compact size, optimized power for onboard operation, and a flexible range of interfaces (CAN, USB, and Mini PCIe).

For applications that require more data collection and processing with real-time responsiveness, the RCX-1440R Advanced Autoware platform provides workstation-class functionality based on the Intel Core i7-97001 processor and Intel C246 chipset. With a custom backplane, the RCX-1440R can support a wider range of interfaces with a maximum power budget of 300 W.

For the most compute-intensive applications, Vecow offers the GPC-1000 Professional platform, which brings real-time AI computing to the edge. With the Intel Xeon E-2278GE processor equipped with the Intel C246 chipset, the GPC-100 offers a maximum power budget of 1500 W. This is enough to support multiple GPUs and enables developers to easily connect and utilize GPU resources for video and AI processing applications.

All of these systems are optimized for in-vehicle operation, supporting a wide range of input power, surge protection, and ignition power control functions. Each system supports fanless operation over the industrial temperature range and is EN50155 certified with shock and vibration protection.

These systems are also designed with flexible interfaces. Instead of building a custom platform, developers can configure these optimized systems to use the sensors and expansion cards of their choice. This also helps future-proof the systems so they can support new sensors, cameras, and other subsystems as these devices become available.

Building blocks for ADAS and autonomous driving

These three systems provide a solid foundation for building ADAS and autonomous driving systems. Choosing between the three options simplifies the evaluation process for manufacturers when determining which platform to use. Since these platforms integrate a lot of software with the hardware, the evaluation is further accelerated. In addition, the availability of three platforms means that manufacturers can choose to migrate to a more powerful platform as their autonomous vehicle applications increase in functionality and complexity.

Typically, ADAS and autonomous driving development is done locally but needs to be delivered to global markets. Various regions (US, UK, Spain, Sweden, Japan, China, etc.) have different regulations, so autonomous vehicle development and testing platforms must be able to reflect these regional differences. Using a consistent industrial computing platform simplifies the process of adapting solutions across regions, markets, and applications. For example, the flexibility of the platform enables developers to use appropriate sensors and cameras for the target region and adapt to new sensors and cameras.

To successfully develop safe and robust self-driving cars, manufacturers must disrupt the way they design cars. Designing hardware requires a high level of expertise to provide enough processing power while safely and reliably meeting the size and power constraints of the vehicle. Many computer platform companies focus on more general markets, and to meet the requirements of the automotive market, they must customize general platforms for each application.

Vecow was able to create an industrial computing platform optimized for the ADAS and autonomous driving industries because of its extensive work in the automotive space and by listening carefully to its customers. For example, by spending a lot of time studying the internal power requirements of GPUs and how to meet those power needs reliably and robustly, Vecow understands the range of functions these systems need to support and how to power them safely.