As the complexity of perception workloads increases, traditional ADAS hardware platforms struggle to meet the growing demand for real-time inference, energy efficiency, and functional safety. This session explores how next-generation automotive System-on-Chips, purpose-built for AI, are redefining edge compute capabilities in ADAS—delivering scalable perception performance while minimizing latency, thermal load, and cost. Drawing on recent use cases and benchmarks, this talk will unpack how ML-optimized SoCs are unlocking new frontiers for sensor fusion, object detection, and situational awareness in L2+ and L3 vehicles—without compromising on efficiency or safety.
In this session, you will learn:
- What makes an SoC “ADAS-ready”? Exploring architecture tradeoffs in ML acceleration, power efficiency, and determinism
- Optimizing AI models for real-time perception on constrained automotive edge platforms
- Integrating ML SoCs into existing E/E architectures: challenges, learnings, and performance metrics
- Opportunities and pitfalls in transitioning from centralized to distributed compute paradigms in ADAS
&nsbp;
powered by Borlabs Cookie