A PickNik developed control stack refined under NASA and Space Force SBIR contracts is cited by the agency in BMW, AWS, Lightspeed, and Hivebotics deployments, with autonomy work pointed at Artemis and commercial stations.
The math problem at the heart of the International Space Station is a clock problem. Astronauts spend roughly one-third of their available crew time moving cargo: unpacking, stowing, transferring bags between modules. That ratio is the constraint that drove NASA's Johnson Space Center to invest in a generation of dexterous robotic arms, and it is the same constraint now shaping what the agency wants from robots on the Moon and future commercial stations.
The software that came out of that investment, called MoveIt Pro, was developed in partnership between NASA's Dexterous Robotics team, led by Shaun Azimi, and PickNik Inc., a Boulder, Colorado engineering firm. The system was released commercially in 2023, and is now appearing in industrial settings NASA describes as directly downstream of the orbital work.
Testing took place in a new Johnson facility, the Integrated Mobile Evaluation Testbed for Robotics Operations, or iMETRO, a lab designed to simulate the constrained choreography of a spacecraft hatch or cargo bag. A NASA Spinoff article on the program describes a sequence in which an astronaut works with a robotic arm to open a hatch or hand off a cargo bag to a free-flying assistant, an interaction that has become routine on station but only after years of tuning.
The funding path runs through NASA's Small Business Innovation Research program and parallel U.S. Space Force contracts, with later SBIR awards refining the user interface, autonomy, and failure-recovery. Ezra Brooks, a principal software engineer at PickNik, previously worked on NASA's now-cancelled On-orbit Servicing, Assembly, and Manufacturing mission, OSAM-1, before joining the company. The chain is laid out on the NASA tech-transfer spinoffs page.
What is new is the reach. NASA says the same software family now appears in BMW's robotic assembly lines, in Amazon Web Services learning-center lobby arm demos, in Lightspeed's housing-panel robots, and in Hivebotics' Singapore-built restroom cleaning robot, called Abluo. PickNik, on its own site, lists additional commercial users: Rapid Robotics, which uses MoveIt Pro in an industrial humanoid called 3Pro, plus CleanBotix and Charles River Analytics. The PickNik product and customer pages describe more than 200 pre-built manipulation skills and a hardware-agnostic posture built on the Robot Operating System, the open-source middleware widely used in robotics research and increasingly in light-industrial settings.
The economic logic cuts both ways. For NASA, paying a small company to harden a software stack that the agency then licenses back is cheaper than building a custom in-house product. For PickNik, SBIR money funds the long, unglamorous work of turning a research prototype into a product that can survive a factory floor. For the buyers, NASA's pitch is faster integration: a buyer can drop in pre-built manipulation skills rather than build motion planning from scratch.
The next test is bigger than cargo. NASA envisions crewed Artemis surface missions lasting roughly a month a year, with the remaining eleven months at Gateway or any near-rectilinear halo orbit operating uncrewed. The agency wants the robots to keep running logistics and maintenance between visits. The same role is expected on commercial low-Earth-orbit stations that will be part-time crewed. MoveIt Pro, per NASA, is one of the candidate control stacks for that workload.
Two caveats matter. The BMW, AWS, Lightspeed, and Hivebotics deployments are described by NASA rather than independently confirmed in this reporting, and the agency has a structural reason to overstate the size of its spinoff footprint. The PickNik "35-person company" figure comes from a Spinoff interview with Brooks, not a current headcount disclosure. Both details are worth a second look before any of these names are cited as fact rather than as NASA-attributed or vendor-attributed claims.
The open question is whether software refined for the slow, careful choreography of orbital cargo will hold up at terrestrial speeds, where BMW's line cycles, Hivebotics' cleaning routes, and an Amazon lobby demo all run under very different real-time pressure than a hatch. The spinoff chain is real. The proof point is the factory floor, not the lab.