Efficient System-on-Chip (SOC) for Automated Driving with High Safety

The requirements of SOCs for autonomous driving systems are described. Huge data from several sensors such as a high-dynamic-range camera, RADAR, and LiDAR have to be processed efficiently to detect, recognize and track objects such as pedestrians and vehicles. Accelerators including a deep neural network are suitable for energy- and area- efficiency. Complex control functions such as path planning, vehicle control and error handling are important to keep the vehicle stable. Multi-core processors and DSPs can perform these function and have the capability for update features. The energy- and area- efficiencies must be improved continuously as well as operating speeds to achieve more safety. The safety features detect faults and prevent fault propagation, and have to meet standards like ISO 26262. The security features are for black-box protection and prevent external attacks. The forecast is concerned with technology progress, the evolution of standards, with multi-sensor inputs and with safe, networked autonomous mobility.