Automated Driving

L2-L3 Automated Vehicles

The Society of Automotive Engineers have defined the levels of driving automation as such:

To address the functionality needed at each of the levels of driving automation, OEMs and tier-1 suppliers need a scalable and reliable architecture to design their AD systems upon. A critical component of AD systems is a domain controller that integrates and processes the massive amount of sensor data required for ADAS and AD features. Examples of critical features for low-level ADAS and high-level AD include:

• Blind Spot Detection (BSD)
• Lane Departure Warning (LDW)
• Adaptive Cruise Control (ACC)
• Automatic Emergency Braking (AEB)
• Lane Keep Assist (LKA)
• Automated Park Assist (APA)
• Traffic Jam Assist (TJA)
• Automated Lane Change (ALC)

L4-L5 Autonomous Robotaxis and Robotrucks

Fully autonomous vehicles including robotaxis and robotrucks require extremely reliable, high performance, power-efficient, adaptive computing. Adaptive XA SoCs provide the optimal balance of hardware accelerators tightly coupled with application software for integrated sensor data aggregation, compute acceleration, and scalar processing.

Data Aggregation

• Programmable I/O to Scale
• Sensor Interfacing
• Native Support for Various Standards
• Data Collection from Disparate Sensors

Pre-Processing

• Data Conditioning
• Sensor Data Synchronization
• Sensor Fusion
• Warping
• Object Classification

Distribution

• Programmable I/O with High-Speed Data Transfer
• Support for a Variety of Inter-Device Connectivity

The data aggregation, pre-processing, and distribution (DAPD) role within a domain controller requires a heterogeneous set of processing engines to process the incoming sensor data – that role can be fulfilled by the adaptive AMD Zynq™ UltraScale+™ XA MPSoC and Versal™ AI Edge platforms. In DAPD, the adaptive XA SoC prepares the incoming sensor data to be processed and then distributed to other elements within the domain controller.

In a compute acceleration role of the domain controller, a key focus for OEMs and robotaxi developers is power-efficient, high-utilization ML inference. Whether it's traditional CV acceleration, CNN processing, or ML acceleration – each of these require an efficient use of TOPs for sensor data processing. The adaptive XA SoC platforms – especially Versal AI Edge – provide the highest AI performance/watt, delivering optimal compute performance. For more on our advantages in compute acceleration.