How to Get Started Developing for Automated Driving

From safe roads to enjoyable commutes, automated driving is poised to change lives and society for the better.

As the car moves to the center of the autonomous world, developers will be tasked with creating innovative and seamless solutions to swiftly respond to and grow with market demands. This requires some serious resources in both the vehicle and data center. Intel already has an ecosystem built with you in mind. With these tools, you'll be able to create—and recreate—the experience of driving.

Automotive basics

Automated Driving Levels

Highly automated driving (HAD): Supports the driver with advanced driver assistance systems (ADAS). These systems entail navigation, safety, security monitoring, image recognition and processing, and sensor data—the data processing hub.

Fully automated driving (FAD): Puts the driver in the passenger seat with onboard intelligence.

Data Center

Data generated by a single automated car equals the data generated by nearly 3,000 people. The data center will be crucial for storing, sharing, and protecting the enormous volume of data generated from deep learning algorithms and keeping the automated vehicle on the road.

5G Connectivity

More than just raw speed, 5G connectivity delivers ultra-low latency at gigabit speeds and high bandwidth. This allows for intelligent and agile networks to give priority to the safety-critical devices required by the automated vehicle.

Human Machine Interface

A software-defined cockpit is the consolidation of cluster displays with in-vehicle infotainment (IVI) systems. This approach seamlessly merges IoT-connected experiences, both inside and outside the vehicle, into a centralized communications, command, and control console that automatically conforms to the needs of individual drivers. The software-defined cockpit is becoming the one-stop shop for:

  • Media management
  • Ubiquitous availability and connectivity
  • Security systems and cloud connections
  • Human-machine interface (HMI) designs that build trust between driver and vehicle

In-Vehicle Computing

As the automated vehicle evolves, it will rely more on sensors, data, and processing power. A vehicle generates nearly one gigabyte of sensor data per second. Intel's portfolio of power-efficient processors, field-programmable gate arrays (FGPAs), and software to make it all work are designed to deliver high-compute performance per watt.

Solutions for Automated Driving

Create an Accident-Free Experience

The complexity of a car learning and navigating through its environment takes more than raw processor power. For a car to sense, learn, and make proper decisions, it needs deep learning algorithms and ways to observe its surroundings.

Building a safe and autonomous vehicle requires:

  1. A deep learning foundation for advanced driver assistance systems (ADAS) algorithms.
  2. The ability to process enormous amounts of environmental data.
  3. The ability to sense surroundings and to:
    • Access hardware accelerators to develop common computer vision routines
    • Provide inference data gathered from cameras
    • Extract features from video and track them

Start Developing for Autonomous Driving

The Intel® Automated Driving SDK (Intel® AD SDK) provides a rich, comprehensive toolkit to build high-performing, power-efficient designs for in-vehicle and cloud-based data center platforms. Data scientists, system designers, and developers of autonomous driving solutions can use this SDK to maximize hardware performance, optimize systems and applications, and advance perception sensor and deep learning algorithms. The SDK includes several workflow modules and optimization tools, including specialized ones for automotive development. It can also be customized so developers can download only what they need.

Learn More

Note:  The SDK is currently available only to existing automated driving customers who are collaborating with Intel. Developers may request access to the SDK. While not all of the tools are publicly available, most are and can be downloaded individually if needed (for example: Intel® System Studio, Intel® Distribution for Python, Free Intel® Performance Libraries, and Intel® Distribution of OpenVINO™ toolkit for the Deep Learning Deployment Toolkit.) Intel is also working on certification to meet the ISO 26262 standard for applications that require Functional Safety (FuSa) certification.¹

In-Vehicle Development

  • Connect the vehicle to the driver
  • Accelerate sensor fusion and environment modeling
  • Optimize performance, tune and debug code
  • Speed up system bring-up and validation

The automated car needs to interact with the driver in the cockpit and make sense of the data it gathers from other sensors. This data also needs to be processed in a timely and efficient manner. The Intel AD SDK tools for in-vehicle software development help create, debug, analyze, and tune code, as well as optimize system bring-up and automate validation. When developing code that runs on in-vehicle hardware, you install the SDK tools on an Intel® processor-based workstation (the host). Code is compiled on the host, which then connects and deposits the code into the in-vehicle hardware (the target).

This development workflow includes the following tools for core software development:

 

Data Center Development

  • Build scalable, multinode applications to manage fleet data and facilitate machine learning
  • Enhance performance, data processing, and more with data center software performance tools

An automated vehicle generates, consumes, and processes an enormous amount of data. You will need tools to accelerate and optimize data processing and to connect to data centers. The data center development tools in the SDK include specific performance optimization libraries and analyzers, plus the same tools as used in in-vehicle development (excluding the system debugger and energy analysis profiler):

  • Intel® Advisor: Use this set of analysis tools for vectorization optimization and thread prototyping.
  • Intel® MPI Library: This library is used to accelerate data center performance for driving simulations and distributed computing analytics for processing vehicle data.
  • Intel® Trace Analyzer and Collector: This analysis tool allows you to visualize data and profile the load balance in your data center.
  • Intel® Distribution for Python*: This distribution is integral for improving code performance in Intel architecture-based data center platforms, particularly for automated driving simulations.

Additional Autonomous Driving Workflows

Other workflows and tools inside the Intel AD SDK help expedite deep learning, data labeling, and visualization, and enable taking full advantage of Intel® FPGAs. These include:

  • FPGA Development: The Intel® FPGA SDK for OpenCL™ software technology eases abstracting for the complexities of FPGA design and writing hardware-accelerated kernel functions in OpenCL™ applications.
  • Deep Learning Deployment: Optimize deep learning models for deployment on autonomous vehicles, and then integrate your deployed model into your autonomous driving application.

Explore the FeaturesGet Started

¹Road Map Notice: All information provided here is subject to change without notice. Contact your Intel representative to obtain the latest Intel product specifications and road maps.

For more complete information about compiler optimizations, see our Optimization Notice.