Visit a classic-car show and you’re guaranteed to see groups of onlookers gathered around the shiny vehicles, asking their proud owners, “So, what’ve you got under the hood?” Ask a gearhead that question, and you’d better settle in for a long answer.
But walk into a distribution center that’s buzzing with autonomous mobile robots (AMRs) and ask the same question, and you’d probably get a quick shrug in response, even as the robot drove itself off to fetch a specific tote a hundred yards away.
So what makes AMRs work? We asked robot vendors, electronics suppliers, and industry analysts that question. They told us it all comes down to motors, software, and sensors.
One of the first things you learn when you look closely at an AMR is that these machines are not built for speed, with zero-to-60 acceleration. Rather, warehouse robots are designed to be slow and steady, toting inventory from point A to point B at a moderate pace, while avoiding collisions with racks, forklifts, and, above all, pedestrians. Safety is job one, so most loaded AMRs—like Locus Robotics’ Origin bot—cruise along at a poky 2.5 mph.
As for their payload capacity, some models have more muscle than others. For example, the capacity of Geek Plus’s P40 model tops out at 88 pounds (40 kilograms), while the company’s P1200 series AMRs can handle loads of up to 2,645 pounds (1,200 kilograms).
When it comes to the “power plant” that enables all this activity, most AMRs run on one or two rechargeable lithium-ion batteries. These are essentially heavier, higher-voltage versions of the lithium-ion batteries used in most smartphones—which makes sense when you consider that AMRs need plenty of juice in order to carry payloads, power their sensors, and hold enough charge to run for an eight-hour shift. And when their batteries get low, the AMRs will steer themselves over to a charging station for a quick refresh—known as “opportunity charging”—or, alternatively, charge up all at once overnight or during a lengthy break between shifts.
That ability to monitor their own power levels comes courtesy of another critical component found inside an AMR: its onboard computer. In addition to monitoring power levels, that computer communicates with the facility’s warehouse management system (WMS) or other software platform, retrieving the AMR’s “marching orders” and enabling the robot to report back again when it has completed each task. These computers typically connect to the facility’s wireless network via Wi-Fi antennas on the AMR.
At the same time, AMRs manage other calculations directly on the vehicle itself, without a wireless link. These tend to be safety-critical calculations like those associated with navigation or collision avoidance, where systems can’t run the risk of losing connectivity through a “dropped call” or a power outage.
If you continue to poke around under the hood of an AMR, you’ll find there’s more technology inside than just the battery, computer, and antennas. There’s also an impressive array of sensors, which essentially act as the machine’s eyes and ears.
These sensors vary greatly in sophistication. On a simpler machine like an automated guided vehicle (AGV), the sensors will probably be inexpensive units—ones that rely on external infrastructure for navigation. In other words, these are sensors that require some sort of “street signs” to give them their bearings, whether it’s infrared beacons on racks and walls, stripes painted on the floor, or quick-response (QR) codes that identify specific locations.
On a more sophisticated vehicle like an AMR, the sensors will likely be higher-end units that allow for autonomous navigation, says Kent Kjaer, sales engineer at the Danish AMR developer Mobile Industrial Robots ApS (MiR). “An AMR won’t just stop dead if it detects a hand truck on the floor but will route [itself] around it, or even back up and take another route. So it needs LIDAR scanners and 3D cameras,” he says.
A LIDAR (light detection and ranging) sensor sees the world in a two-dimensional (2D) plane at a height of about eight inches off the ground. That’s enough to allow a robot to determine where it is in relation to its physical surroundings, including walls, doors, and racks—a process known as localization. But it’s not enough to provide a failsafe collision-avoidance solution: The robot still might miss a pallet lying on the floor or a pair of lift-truck forks extended above its detection range.
So AMR developers typically add another sensor called a three-dimensional (3D) camera that maps the robot’s surroundings roughly from floor level up to a height of five or six feet. MiR, for example, combines a forward-looking 3D camera with a 270-degree field of view (FOV) with a similar one looking backward, and combines the two inputs for a full, 360-degree picture, Kjaer says.
MiR also adds proximity sensors in each corner of the AMR to detect any nearby objects that the cameras may have missed (like an object someone just placed on the ground). As an AMR moves through its environment, it fuses those multiple sensor inputs into a single image that refreshes many times per second, helping it avoid collisions and find its way through a process called simultaneous localization and mapping (SLAM).
Compared to years past, today’s engineers have an unprecedented array of specialized sensors to choose from when designing an AMR, says Tyler Glieden, a market product manager at SICK Inc., a German sensor manufacturer.
For example, they can opt for a mechanical LIDAR sensor, which works by shooting a laser beam that reflects off a spinning mirror in different directions, and then measures the time of flight (TOF) as the laser light reflects off its surroundings. Newer models achieve the same end with solid-state LIDAR that works without moving parts. Either way, as a robot rolls faster, it can adjust those sensors to “look” farther down the road, giving it more advance warning of obstacles and ensuring it has enough time to stop even while carrying a heavy load.
They also have their pick of LIDAR sensors that are rated for outdoor use—meaning they allow a robot that moves between buildings to navigate in low-visibility conditions, like rain, snow, and fog. Still other models are designed for robots that work in refrigerated storage areas and freezers, with features that prevent them from fogging up as the temperature changes.
To manage all those variables, SICK also offers the “Sick Safety System,” which collects laser scanner data and analyzes it using software on board the AMR. That approach helps the robot avoid collisions and downtime by adjusting its driving speed to the situation.
As AMR applications evolve, robot developers continue to gussy up their offerings with new sensors and features. Some robots now incorporate digital cameras that let them take pictures of bar codes and inventory, while others feature “height sensors” that help them fetch totes off high shelves. Still other models come equipped with the equivalent of an automobile’s headlights and horn, enabling them to honk and flash before turning a corner to alert warehouse workers to their presence.And there’s no indication that this R&D work will stop anytime soon. AMR manufacturers will continue “souping up” their robots with improved sensors, batteries, and computers, creating virtual muscle cars that can keep pace with changing logistics demands.