FedEx's two-armed DexR robot loads delivery trucks like 3D Tetris

FedEx handles more than 15 million packages a day, and one of the trickiest jobs falls to humans: loading a delivery truck with boxes of every size, shape, weight and packaging material, arriving randomized. In September 2023 the company unveiled DexR, a two-armed robot built with Redwood City robotics startup Dexterity, that automates exactly that step — stacking rows of differently sized boxes inside a truck as efficiently as possible.

The robot perceives packages with cameras and lidar, then plans how to configure the available boxes into a neat wall, placing them snugly without crushing anything. Each placement is confirmed by force feedback, and after every box the system rescans the stack with cameras and depth sensors to compare it against its model; any discrepancy forces the stacking plan to adapt on the fly. Dexterity CEO Samir Menon says the robot uses generative AI both to work out the stacking and to identify and grasp boxes, with the systems woven together by careful engineering.

'A few years ago, AI was not at a stage where it was smart enough to handle this kind of complex decision-making', said Rebecca Yeung, FedEx's vice president of operations and advanced technology. DexR was in testing ahead of a wider FedEx rollout, part of a wave of commercial robotics that Carnegie Mellon's Matthew Johnson-Roberson expects to grow as AI advances finally make business cases in cost and efficiency.

Teach a robot to build 'walls' of mixed boxes like a 3D Tetris player: perceive, plan the whole stack, seat each box with force feedback, and re-plan when reality drifts.

Loading rewards whole-stack planning, not repetitive motion — the opposite of what industrial robots were built for.

Simulation-trained AI breaks when it meets real cardboard; DexR closes the loop by re-scanning reality after every placement.

Randomized, crushing-sensitive inputs make loading a genuine optimization problem rather than a pick-and-place one.

FedEx targeted the specific step where human labor is hardest to keep, instead of trying to automate everything at once.

The hard part of warehouse robotics is not gripping — it is planning a whole arrangement under real-world drift, then verifying every placement against the plan.

As of the unveiling, DexR was in testing ahead of a wider rollout across FedEx operations; WIRED framed it as evidence that advances in AI, sensors and hardware were opening commercial robotics applications beyond the factory line.