The automotive industry is rapidly converging on centralized compute architectures and AI-driven perception stacks. However, scaling from L2 ADAS to higher levels of autonomy is not a linear exercise in adding compute or adopting new models. This keynote examines how perception architectures, sensing strategies, and compute topologies evolve across levels of autonomy. It challenges the binary framing of centralized vs. distributed systems by identifying which functions naturally consolidate – and which make sense to remain at the edge. This talk connects sensing modalities, perception strategies, and interior sensing approaches with architectural design patterns against the backdrop of real world scalability constraints.
