Willow Garage

» see willowgarage.com as it lived, on the wayback machine

~ the obituary ~

willow garage sat next to the papers and repositories i kept reading through the mid-2010s. i do computer vision on small airborne platforms in ankara. opencv is a library i have imported roughly a thousand times. ros is a runtime i have run on hardware more than once. both of those things came out of a single menlo park lab that closed before most people who use its outputs knew it existed. either way, that is the shape of the willow garage death.

of my life, but not by much. scott hassan had co-written the original google search backend as a stanford grad student, then built the egroups service and sold it to yahoo. by the mid-2000s he had enough personal capital to fund a research lab on the same self-directed terms bell labs had once operated on. willow garage opened in late 2006 in menlo park with steve cousins as ceo. the mission statement was "impactful and non-obvious" personal robotics.

the pr2 was the flagship hardware, a two-armed humanoid research platform with 7-DOF arms and enough sensors to be dangerous. by summer '09 a pr2 had opened doors, located a wall outlet, and plugged itself in unsupervised. in may 2010 willow garage held a graduation event and shipped eleven pr2 units to research groups at berkeley, stanford, mit, tum, and the rest. that was the piece the mainstream press covered. it was not the piece that mattered longest.

~ the case file ~

~ the ROS moment ~

ros 1.0 shipped in january 2010. the robot operating system is not an operating system in the linux sense. it is a message-passing framework with a package layout convention on top, letting a perception node and a motor-control node run on separate processes and talk to each other over a well-defined bus. before ros, every research lab wrote its own middleware from scratch. after ros, a graduate student could pull a working stack from apt-get and be steering a mobile base inside an afternoon.

opencv was already a decade old when willow garage inherited it from intel around 2008. the lab put paid maintainers on it and ran the api cleanup that made the c++ interface tolerable to actually use. it also shipped the point cloud library alongside for 3D data. those three libraries plus the pr2's ikea-tier reference implementation are the reason a small drone project in ankara in 2026 can bootstrap a computer-vision loop in a weekend instead of a year. that piece of the story genuinely reached me. it does not get counted in the lab's balance-sheet outcomes.

~ what really happened ~

willow garage did not fail in a market sense. it was never trying to sell anything at scale. it was scott hassan's personal research budget structured as a company. in early '14 hassan decided he was done funding it and the lab wound down. the pr2 hardware line moved to a spinoff called suitable technologies, which had already been separated out to sell the beam telepresence robot. the ros maintenance moved to the open source robotics foundation, later reorganised as open robotics. the willowgarage.com website stayed online as a redirect through april 2021 and then finally lapsed.

i saw this at the time as an odd shutdown. the lab was doing better technical work than most well-funded startups in its category. what actually happened was that the private-lab model requires the private funder to keep wanting the lab to exist, and hassan stopped wanting that. the pattern shows up again with bell labs, xerox parc, and every other research organisation that has ever depended on one patron. willow garage was much smaller and much shorter-lived, the mechanism was the same.

~ what got buried with it ~

in fairness, willow garage's alumni network is one of the most productive diasporas in the industry. open robotics still runs ros as a foundation-hosted project. fetch robotics ended up inside zebra after a bumpy warehouse-automation ride, savioke shipped hotel-delivery robots into hilton properties for years before it wound down. zipline moved medical supplies by drone across rwanda and later ghana, and redwood robotics went to google. the pr2 hardware itself sold in the low tens of units and none of it ships anymore. the software the lab funded ships on hundreds of thousands of robots, most of which do not know willow garage was the reason.

the lab existed for about seven years. in that window it published foundational software that outlived it by more than a decade and is still the default choice in its category. the willow garage grave is unusual for the graveyard, most entries here died because their users left or their money ran out in a way that killed the thing they made. willow garage's outputs did not die. only the address did.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »