Level 5 Autonomy has Arrived at the Warehouse

Level 5 Autonomy has Arrived at the Warehouse

In the world of self-driving cars, Level 5 autonomy—which indicates full autonomy in all environments, all the time—is considered the Holy Grail. Over the past decade, the industry has been inching closer and closer to that goal, but drivers are still waiting to let go of the wheel and spend their commute time doing anything but driving a car. Last week, Alphabet’s Waymo and GM’s Cruise received preliminary approval to offer autonomous ride-hailing car services in California. It’s not a done deal (approval from the California Public Utilities Commission remains pending), and Waymo’s service includes the presence of a ‘safety driver,’ but it’s progress. Level 5 autonomy is slowly but surely becoming a reality. At the moment, however, the benefits of Level 5 autonomy seem to be just out of reach—at least in the world of self-driving cars.

Inside the warehouse, the story is completely different.

At ETH Zürich, my team has been exploring and redefining the capabilities of autonomous systems for the past 14 years. With a focus on developing fully autonomous self-flying drones, our challenges were certainly less dramatic than those faced by the auto industry—no lives were on the line. And unlike automobiles, our solution was not restricted by such weighty barriers to entry as international regulations, adherence to road signage, or the need to operate in all types of weather. Still, delivering a system capable of operating indoors safely, reliably, and with complete autonomy was no easy task.

It helped that we had a significant head start thanks to the work we did back in the Kiva days when we created the first Level 5 autonomous mobile robots (AMRs) for ‘goods to man’ product movement in the warehouse. However, unlike the Kiva robots, which enjoy the luxury of simply stopping and powering down when they come near a human worker or a forklift, drones must be capable of detecting and avoiding obstacles in flight and, in the case of an operational failure, of returning to a charger rather than crashing to the ground. Drones must also navigate in three dimensions to complete their missions—without the support of GPS. To address these needs, we developed technologies to enable indoor navigation and positioning; robust localization; path planning, obstacle detection, and collision avoidance; and the ability to detect and respond automatically to any abnormal conditions. We also built out sophisticated adaptation and learning capabilities to ensure the drones could continue to perform well over the long term—even when their sensors and actuators aged and changed over time. And to give the drones the ability to stay powered for the duration of each mission, we developed technology that enables autonomous battery charging in between flights.

These individual components were important, but just as critical to achieving our own Holy Grail of Level 5 autonomy in the warehouse was the system itself. Unless the drones could share information among themselves and with warehouse management systems (WMS), the system would still require a human operator. To address this piece of the puzzle, we developed tools that enable the drones to communicate and coordinate as a fleet. This ensures there are no mid-air collisions, that work is allocated equally so each drone completes its fair share of inventory checks, and that one drone can take over in-process tasks from another drone when needed. If the system determines it’s necessary, a drone can even be assigned to double check the work of another drone co-worker. And not only does the system share information and insights with the WMS after each mission, but it also receives information from the WMS to automatically determine which locations to inventory on any given night.

The results of our research speak for themselves. Our inventory tracking system powered by self-flying drones was not only the first fully autonomous system on the market, but it remains the only system of its kind today. Our clients rave about the ability to track inventory without any human intervention—a costly, labor-intensive task that is tedious, dangerous, and famously dreaded among warehouse workers. The only thing required of warehouse personnel is to turn the system on at the start of a cycle, and then turn it off when they return to work the following day. While employees are at home sleeping or enjoying a day off, the system is hard at work, collecting accurate data about what inventory is where, identifying inventory that isn’t where it should be, and then transforming that information into insights that are fed straight into the WMS. When the worker returns to the warehouse the next morning, it’s easy to see what’s right, what’s not, and then take action to remedy any issues before they disrupt the business.

I firmly believe Waymo, GM, and the rest of the automotive industry will achieve Level 5 autonomy—someday. Drivers will eventually be able to take their hands from the wheel and turn their attention to more productive activities, even during the most hectic commute. But inside a growing number of warehouses around the globe, Level 5 autonomy for inventory tracking has already arrived. Using our Verity system, warehouse employees are able to check the task of inventory tracking off their to-do lists and focus on bringing greater value to the business by doing the tasks that only humans can do.