Perception is foundational to enabling advanced capabilities such as navigation, SLAM, and AI in modern robots and autonomous vehicles. Yet when sensors stop working, robots stop working, too. As a result, today’s perception systems and the sensors they rely on are under increased scrutiny for capability and reliability — particularly when fleets scale into the hundreds or thousands of devices. In this talk, we’ll explore how the underlying infrastructure of perception — namely sensor calibration, sensor fusion, and sensor monitoring agents — can be used to increase uptime, unblock performance, and reveal aggregate health trends for large fleets of deployed devices.