UPS and FedEx anticipate delivering close to 1 billion packages in the next few weeks. That's a 10 percent uptick from last year and, if accurate, a new shipping record and high water mark for American consumerism. USA! USA! USA!
Moving that much merch quickly and accurately would be impossible without the recent automation revolution in the logistics industry. Robotics technology developed by Kiva has enabled Amazon to fulfill orders same-day in many locations. Though human workers still play a vital (albeit controversial, according to recent reports) role in the picking, packing, and palletizing that go into dispatching goods to your doorstep, there can be no question that we're getting ever-closer to a so-called lights-out shipping warehouse, one in which all the workers are robots.
One recent automation breakthrough (or job casualties, depending on your slant) is vision guided vehicles (VGV).
"Our robots run quite nicely in parallel with humans," said Jim Rock, CEO of Seegrid, the biggest supplier of vision-guided vehicles to industry. "We can show up, drop off one of our robots, and in a few minutes it will be moving around a warehouse without any need for additional infrastructure."
The working around humans part is important. Temporarily, at least, we're in a transition period during which humans and robots will work side by side at the fulfillment centers responsible for delivering your last-minute Christmas shopping. Speed is critical, but so is safety, which is why autonomous robots with 3D vision are such a promising development.
Seegrid's autonomous forklifts zip around carrying heavy pallets crammed with goods from trucking bays to robotic workstations, where the merchandise will be sorted for delivery. They can do this quickly and with far greater safety than human-operated forklifts, which are a menace to worker safety. According to Tools Of The Trade, there are 110,000 forklift accidents in the US every year and more than 100 deaths.
Seegrid's autonomous vehicles rely on the convergence of two technologies: 3D sensing, which has become affordable in the last few years thanks to cheap cameras and increasing computing power, and evidence grid mapping, a newer strategy that enables autonomous navigation on the fly. Instead of registering objects, which is difficult to do in real time, robotic systems using evidence grid maps accumulate occupancy evidence for an array of spatial locations, slowly resolving ambiguities as the robot moves. (If you yearn for a more detailed explanation, check out this seminal paper, which introduced the idea to the world.)
"We're buying very inexpensive cameras off the shelf," says Rock. "That's the least interesting part of what we do. It's the evidence grid that's very very complicated and that frankly took our engineering team decades to work out."
He's not exaggerating. The man helming Seegrid's engineering efforts is Hans Moravec, a professor of robotics at Carnegie Mellon University and the father of evidence grid maps. One of Moravec's pupils at CMU was Chris Urmson, who is now the director of Google's self-driving car program.
Seegrid's vehicles are one example of the new flexible automation that's allowing fulfillment centers to rapidly reconfigure and better absorb peak capacity spikes around the holidays. Canadian robot maker ClearPath has a self-driving vehicle for the factory floor, and British design firm Cambridge Consultants recently demonstrated that existing machine vision and machine learning capabilities can be harnessed to enable industrial robots execute tasks that aren't rigidly defined.
We don't have a lights-out fulfillment center yet, but if Americans continue doing their shopping online, there's a good chance that your gifts will pass through even fewer human hands on their way to your doorstep next year.