The field of robotics is about to enter an exponential growth phase. Robots from corporations such as Qinetic, irobot, and Vecna Robotics are already used extensively by the military and industry, and new robotics corporations, such as Seegrid and Heartland robotics, have formed. A combination of factors, including Moore’s law, the increasing cost of human labor, and improved actuators/sensors, mean that robots are beginning to appear in ever greater numbers and increasingly diverse locations. In an interview with Sander Olson, Seegrid CEO Dr. Scott Friedman discusses the burgeoning importance of warehouse robots, why voice-activated industrial robots won’t appear, and the coming era of “twilight” factories and warehouses dominated by robots.
Dr. Scott Friedman
Question: Seegrid’s warehouse robots have vision systems. How does the vision system work?
The vision system works by creating a 3-D grid, actually an Occupancy Grid, which is a machine vision technique invented by Seegrid’s co-founder, Hans Moravec. That is why our company is named Seegrid. It is a pun on “see the grid”.
Question: Can the vision system recognize specific objects?
The system is used to let the robot know where it is located on the grid, what we call localization. It isn’t designed for object recognition, but we can customize the unit and add that capability if our customers desired it. Question: Can these robots work in any standard warehouse? Do the warehouses need to be modified in any way? It is one of our main selling points that warehouse modifications need not be made in order to implement Seegrid Robotic Industrial Trucks into our customers’ processes. The evidence-grid technology that we have developed for location recognition is a general, or strong, AI (Artificial Intelligence), so it doesn’t need specific conditions in order to work reliably; it locates its position regardless of where it is. The system works great outside as well as inside, though the current products are not waterproof, so we don’t recommend using them outside. Question: How long does it take for a robot to learn its way around a warehouse? A few hours. We call our system “WalkThroughThenWork®”. The customer takes the robot around the warehouse once, and shows it what it wants to do. Then, after the walkthrough, the robot starts to work. So, these robots are engaging in productive work on the day that they arrive in the customers’ warehouse. We just had a major soft drink manufacturer go live using our robots—they were up and running the same day. Question: How close are we to “lights out” warehouses run exclusively by robots? I’m not sure that we will ever see warehouses that employ robots exclusively. The ratio of robots to humans will continue to rapidly change until we see “twilight” warehouses with large numbers of robots and maybe a dozen or so humans. In aviation, planes have been able to take off, fly and land for decades, however there is such complexity that some humans still need to be in the loop for traffic control. Warehouses have that same kind of complexity. Question: Will voice-operated robots soon become an option on future robots? Not in an industrial environment. The ambient background noise presents serious problems. It is technically feasible, but for most tasks some sort of tablet-based solution generally makes more sense. A dedicated audio system is expensive, so unless the person needs to make frequent commands to the robots, it isn’t cost-effective. Question: How big is the potential market for warehouse robots? The global market is for perhaps a million robotic units per year; the installed base of industrial vehicles is about 11 million. We have just announced partnerships with the two largest industrial vehicle companies in the world. The largest industrial vehicle maker in the world is Toyota, and we announced a partnership with them in May. A couple of days later, we announced a partnership with the second largest industrial vehicle maker, Linde Material Handling. Both of these partners are producing robotic vehicles which incorporate Seegrid technology; we call it “Guided by Seegrid”. Question: What percent of the cost of running a forklift is currently manual labor? Labor is approximately 80% of the costs of operating forklifts, so it is not a matter of if robotic vehicles will become the standard, it is a matter of when. Within a decade, perhaps 30% of all industrial vehicles will be robotic, and the ratio will get continually larger. Question: How safe are these robots, compared to human operated machines? Robotic industrial vehicles are, according to the insurance companies, ten times as safe as human-operated vehicles. In addition to reducing costs, improving safety is another reason that this industry is becoming increasingly interested in robotics.
Question: In Robot: Mere Machine to Transcendent Mind, Seegrid Chief Scientist Hans Moravec argues that limited computing power is one of the main limiters of robots. Do you agree? Yes, right now computing power is one of our main limiters. The capabilities of robots follow Moore’s law fairly closely, so we should have much more capable and versatile robots in the coming decade. More computing power allows us to provide the robots with better vision, more sensors and better AI. Increased computing power will be the big driver of robots during the next decade. Question: Can your robots take advantage of cloud-computing? Can they employ GPUs (Graphical Processing Units)? Cloud computing really isn’t an option with robots. Bandwidth is increasing, but computational use is increasing faster. For instance, our robots used to employ low-resolution cameras. Now they use 10 VGA cameras, so there is no way to provide sufficiently high bandwidth to a server farm. Therefore, the robot’s computing power has to be local. The robotics industry is very interested in GPUs. In fact, robots may become the main consumer of GPUs a decade from now. Question: What capabilities will Seegrid’s next-generation robots have? When will they become available? Over time, robots will be able to operate in an increasingly autonomous manner. They will be able to accept more complex commands, such as “unload these pallets over there, and stack them in those racks”. Robots will become increasingly adept at recognizing people, what they are doing and how best to work alongside them. Question: What industry will be the biggest market for robots during the next decade? Unless personal robots advance substantially, the biggest use of robotics in the next decade will definitely be industrial vehicles. At some point, the personal robotics industry will take off, and there will be a robot in every home. I don’t know if that will happen in five years or twenty, but I am certain that millions of robotic vehicles will be built.