Software Defined Vehicles
Interview with Deepak Vedha Raj Sudhakar, Mercedes-Benz
“Performance simulation and predictive analysis play a critical role”
Balancing performance and functional safety remains one of the toughest challenges in automotive computing. Mercedes-Benz expert Deepak Vedha Raj Sudhakar explains how model-based simulation optimizes SoC design and accelerates ADAS integration.
With a background in embedded systems and timing analysis, Dr Deepak Vedha Raj Sudhakar has spent nearly a decade shaping the future of high-performance automotive computing. Before joining Mercedes-Benz as SoC Performance Architect, he worked as a researcher at Inchron, focusing on timing analysis and optimization of distributed E/E architectures, and as an embedded software engineer at DAF Trucks, where he contributed to vehicle intelligence and control systems. At the Automotive Computing Conference 2025, he will present a model-based approach to predictive resource analysis for early-stage SoC design. In the run-up to the event, we asked him three questions on this topic area.
ADT: Modern vehicle platforms demand high performance and strict safety compliance at the same time. From your perspective, what are the key challenges in balancing SoC performance optimization with functional safety?
Sudhakar: Functional safety introduces strict constraints into the SoC performance optimization process. To guarantee safety, certain compute cores and memory regions must be reserved for specific safety-critical applications. This limits flexibility in resource allocation and reduces the design space for optimization. As a result, performance tuning must be carried out within these boundaries, ensuring that safety requirements are never compromised while still achieving the best possible system efficiency.
At the ACC 2025, you will present an approach for early-stage predictive resource analysis using model-based simulation. How does this method improve development efficiency, especially in integrating ADAS functions?
This approach uses model-based simulation to predict resource usage—CPU, GPU, NPU, interconnect, flash, and RAM—under realistic worst-case end-user scenarios with certain foreground and background applications running simultaneously. By simulating these conditions early, we avoid over-dimensioning hardware and gain deep insights into system behavior. For example, does a safety-relevant domain such as the interior camera for driver and passenger monitoring risk CPU starvation? Are task dispatch rates and run queues appropriate for the CPU? Can the system handle both CPU-bound and memory- or I/O-bound tasks effectively? These insights help identify optimal configurations that meet both safety and performance requirements long before hardware or implementation is available, significantly improving development efficiency and the integration of complex ADAS functions.
Mercedes-Benz is known for pushing technical boundaries while maintaining reliability. How do performance simulation and predictive analysis contribute to achieving both innovation speed and system stability?
Performance simulation and predictive analysis play a critical role in determining how future SoCs should be scaled and what compute and memory resources will be required for upcoming features. These techniques also enable the comparison of different SoC options across vendors, providing valuable insights for selection from a performance perspective. Additionally, predictive analysis helps identify resource headroom early, ensuring that new functionalities can be integrated without compromising system stability or safety.
