The SoC provides a complete embedded solution for Kodiak’s multi-sensor perception, fusion and path planning - running neural networks while processing multiple cameras, lidars and radars.
To date, Kodiak has logged over 300,000 miles on Ambarella’s CV2 AI perception SoC while delivering over 875 loads, demonstrating that Kodiak Driver with help from the Ambarella hardware can withstand the rigorous environment of highway trucking.
The CV2 has enabled premium image quality whether in rain, fog or darkness, ensuring that freight deliveries arrive both safely and on time.
Kodiak’s next-generation self-driving truck platform will utilise the CV3-AD685 SoC to optimise hardware cost, reliability, size and robustness, while improving safety and performance. Designed with Ambarella’s “algorithm first” philosophy, this 5nm SoC is purpose-built to provide a high-performance compute solution in the smallest form factor, which enables Kodiak’s modular autonomous vehicle platform hardware across a wide variety of environmental conditions.
“The CV3-AD685 enables the Kodiak Driver, Kodiak’s autonomous system, to turn sensor data into actionable insight with lower latency, lower power consumption, and higher reliability,” said Don Burnette, Kodiak’s Founder and CEO. “We are excited to exhibit a Kodiak autonomous truck at Ambarella’s invitation-only event during CES for the second year in a row, demonstrating how our partnership furthers the mission of autonomous trucking.”
According to Fermi Wang, President and CEO at Ambarella, “Our CV3-AD685 SoC provides the autonomous vehicle industry with the processing performance needed to set new benchmarks in AI-based autonomous safety, efficiency and innovation.”
The CV3-AD685 AI domain controller offers a combination of highly efficient AI processing, advanced image signal processing and ultra-low power consumption. Its next-generation CVflow AI engine includes a neural vector processor engine that is 20x faster than the previous generation of CV2 SoCs, along with a general vector processor to provide the overall performance needed for full autonomous driving stacks.
