Commercial operations of autonomous fleets of cars are demonstrating how important fusing multiple sensing modalities is for handling edge cases like fast-changing weather and complex urban scenes. Lidars provide perception pipelines with low-latency, 3D point clouds — providing geometrically accurate data that can be fused with camera and radar imagery. The sensor sensitivities degrade differently in various weather conditions, allowing for robust detection of other vehicles and vulnerable road users. Recent trends in detectors and emitters are favorable for Level 4 sensing at low cost. We will also share some examples of how rapid progress in machine learning can be applied to sensor processing and to the understanding of complex scenes.