Search

Autonomous Vehicle What's in an Autonomous Vehicle platform?

| Author / Editor: Seth Lambert / Erika Granath

Autonomous Vehicles (AVs) need an array of components for self-driving. Some companies producing AV technology supply every component that's needed to permit a vehicle to drive from Point A to Point B on its own (what's termed Society of Automotive Engineers [SAE] Autonomy Level 4 or 5 ["attention-free" driving]). Other companies supply only some of these components.

NVIDIA DRIVE software enables key self-driving functionalities such as sensor fusion and perception.
NVIDIA DRIVE software enables key self-driving functionalities such as sensor fusion and perception.
(Source: NVIDIA)

Safety is a top priority for any autonomous vehicle platform, and the layers of AV tech platforms enable safety via ultra-fast, parallel processing under less than ideal performance conditions (heat, vibration, and other negative environmental factors).

Many platforms are modular, allowing automaker OEMs to "plug-and-play," adding their own complement of sensors, operating systems (OSes), network connections, and processing and decision-making software.

What's in an Autonomous Vehicle platform?

Most autonomous vehicle (AV) "stacks," or sets of hardware/software subsystems, include the following six layers:

Sensors

These are the eyes and ears of AV platforms. They both transmit and receive signals in the form of still images, video, radar, sonar, LiDAR, and other technologies in order to convey the real-world environment and obstacles to a car's navigation system. These sensors often come with their own digital signal processing (DSP) capabilities.

Perception Engine

This is where data from a car's sensors is fused and analyzed. Graphics processing and artificial intelligence (AI) algorithms classify objects and determine whether they pose a danger to the vehicle. Over the longer term, data processed by a perception engine can be aggregated in cloud-based storage and compared to data from other vehicles to improve the algorithms utilized cumulatively.

Mapping/Location-Based Services

Maps are either created on the fly or loaded into memory from onboard hardware, so GPS-based location positioning can be done with pinpoint accuracy. Location data can help immensely with navigation when fused with sensor data (see above). Infrastructure and road condition changes can be assimilated so that only the most current environmental data is used.

Processing Center

This is the "brain" of an AV platform where data is crunched — often using parallel system-on-a-chip (SoC) microprocessors and graphic accelerators. This is where the operating system of a platform resides.

Data Connectivity

This hardware sends and receives signals between all the platform components to and from one another. It also manages communications with outside networks via cellular or Bluetooth transmission. Data is compressed, decompressed, encrypted, or unencrypted as necessary.

Safety/Security Gatekeeping

This layer makes sure that instructions processed by the AV stack fall within the safety and operational requirements of the vehicle. Hardware and software here can signal abnormalities, reduce malfunctions, and handle subsystem failures. There are also cybersecurity features that prevent hackers or unauthorized users from taking control of the platform.

While numerous companies tie together all or most of the above pieces in engineered platforms and apply their own control mechanisms, some produce only a select range of each of the above layers. Some firms supply software development kits (SDKs) that allow carmakers to customize and tweak the discrete layers on their own.

(ID:46492520)