Fizyr, the robotic parcel induction company, continues to provide exclusive insights into its latest machine deep learning technology at Parcel+Post Expo 2021.
Exhibiting its product in conjunction with EuroSort’s Split Tray Sorter, Fizyr’s AI demonstrated how the company has been able to automate the last manual link in the picking sortation chain. The ability to detect and pick up a box that is slightly out of place has so far been a uniquely human skill. To keep up with the increase in parcels over the past year, sortation centers would have had to employ a much greater number of people to do these very simple tasks. However, Fizyr’s technology rises to the challenge.
Shubham Singh, technical consultant at Fizyr, explained, “If you move a box just slightly, it becomes an unknown thing to a computer. Yet even a human baby can identify that a box is still a box if it is slightly at an angle, and manipulate it appropriately. Basically, robots are dumb. That’s where our AI comes in – we have built a brain that goes into these robots.
“It understands its environment, the dimensions of a parcel and can decide how to go and pick one up. This is the key missing factor in why this step in the sortation process hasn’t been automated so far. Every year, 20 billion parcels are coming to market. You cannot just scale up workforces this quickly in response. Rather than employing more people or increasing existing shifts from six to eight hours, for example, this increase in parcels can be dealt with by a robot.”
The tech itself is deep learning – training the robot through a replicated neural network with several images. It does not use reinforcement learning, to prevent it from learning the wrong process once installed in the sortation center. At the Parcel+Post Expo showcase, there’s a standard laptop with a GPU sitting underneath the robot, which is processing all the information from the camera in real time.
In his technical presentation on the morning of Day 2 of the show (October 13), Herbert ten Have, CEO of Fizyr, said, “Fizyr’s computer vision software is highly reliable as it is able to detect close to 100% of the parcels in bulk and is already used globally in production. The key challenge that our software addresses is segmentation of overlapping parcels because, unlike other AI solutions, our software can be 100% sure there are two different envelopes. The same applies for boxes with tape – it won’t get fooled into thinking one box is six boxes. AI can also see better than humans can. It can find all boxes even in darker conditions. Finally, when handling delicate packages, our picking solution ensures careful handling with its soft-gripping technology.”
Singh added, “It can also prioritize which box to go and pick up first, and can identify which parcels are damaged – and decide to not pick them up. For example, women’s cosmetics have to be handled very delicately, otherwise they will break. In this case, you can customize how the robot will pick it up.”
Fizyr works with companies like EuroSort as a logistics integrator. It designs the solution and automates the picking part. This means that whichever way up the parcel is presented, Fizyr’s technology should be able to understand what the parcel is, what shape it is and how to pick it up. This comes with the added benefit of reducing contact in sortation centers, which is where much of Fizyr’s most recent client interest has come from.