Aptiv, the auto parts supplier, is putting its radar plus vision PULSE sensor into warehouse robots. Robust.AI's choice of it for the Gen 3 Carter shows how safety rules for human cooperative robots are starting to dictate what those machines need to perceive.
Warehouse robots are running into the limits of cameras. Real fulfillment floors push dust, glare, moisture, reflective shrink-wrap, and forklifts into a single field of view, and a vision-only system can stumble in conditions that arrive in minutes. The result is a quiet but consequential shift in what warehouse robots are being spec'd to see. Instead of one camera and a machine-learning model trying to make sense of every situation, the next generation is being built around sensor fusion that pairs radar with vision and runs machine learning on the combined raw data before any single-sensor interpretation.
That shift is now visible in a supplier win that looks small but lands in a category that does not change often. Robust.AI, the San Carlos, California-based maker of the Carter collaborative mobile robot, has selected Aptiv's PULSE perception system for the third generation of Carter, the company's order-fulfillment and mobile-sorting robot, sold under a robotics-as-a-service (RaaS) contract.
PULSE, which Aptiv describes as a radar-plus-vision sensor with on-device machine learning, is a sign of the auto-parts giant's move from cars into warehouse robotics. Aptiv pitches it as fusing radar and camera data at the raw-signal level, so the model sees an obstacle, not a radar return layered on top of a picture. The pitch is reliability in conditions that have historically punished camera-only systems: dust, glare, moisture, and reflective surfaces. In an Aptiv Q&A on the sensor, the company frames PULSE as a "new take on perception" built around ML on raw sensor data rather than cascaded perception pipelines.
The framing Aptiv leans on hardest is functional safety. In a coordinated announcement echoed across trade and wire coverage, Jay Bellissimo, Aptiv's senior vice president of intelligent systems and president of software and services, ties the Carter selection to the company's path toward functional-safety compliance for physical AI. That is a real and specific claim, and it is the part most worth watching. Collaborative robots, by definition, share workspace with humans, and the engineering bar for fail-safe perception is rising. Buying a sensor stack that an automaker already validates against ISO 26262-style safety cases is, for a warehouse-robot vendor, a way to inherit a safety story instead of writing one from scratch.
One customer perspective anchors the choice. Anthony Jules, Robust.AI's co-founder and CEO, frames PULSE as a differentiated camera-and-radar approach that the company expects to translate into uptime on floors where camera-only peers have struggled. The corroborated coverage across Robot Report, Stock Titan, and Robotics Tomorrow does not add independent uptime numbers. The case for radar-vision fusion therefore rests on Aptiv's stated mechanism and the company's auto-industry safety record, not on head-to-head warehouse data.
The RaaS part of the story is doing more work than the press release acknowledges. Robust.AI sells Carter under a performance-based contract, which means the vendor is on the hook for the robot's productivity, not just the hardware. Under that model, a perception failure is not a warranty claim; it is a missed pick, a delayed tote, and a contractual service-level credit. Spec'ing a sensor stack that survives a wider envelope of warehouse conditions is, in RaaS economics, the cheaper insurance. The RaaS design is also what gives Aptiv a wedge: warehouses that would never sign a capital-expenditure purchase order for a premium perception system can be exposed to it through the robot vendor's per-order pricing.
The falsifier is straightforward. If camera-only warehouse robots continue to hit their uptime targets in real fulfillment conditions without triggering functional-safety incidents, the urgency argument for radar-vision fusion weakens, and Aptiv's warehouse foothold stays narrow. If they do not, expect the next round of Carter-class announcements to name a fusion partner of their own.
What to watch: a second warehouse-robot vendor adopting radar-vision in the next two product cycles; Aptiv's RaaS pipeline beyond Robust.AI; and whether the functional-safety framing becomes a procurement requirement, not a marketing line.