AI-Driven Robots Handle High-Speed Logistics Sorting and Depalletizing

Chances are that any products you’ve ordered online that arrive on your doorstep in poly bags, boxes, or flat padded envelopes have been touched by a robot. More than ever before, warehouses and logistics centers are relying on advanced automation technologies such as robots, machine vision systems, and artificial intelligence (AI)-based software to automate processes and help keep the flow of goods in motion.

As these facilities face a constantly changing landscape of product variability, their automation systems must be able to handle new package shapes and types while maintaining throughput rates and driving productivity improvements. Parcel-handling robotics company Plus One Robotics (San Antonio, Texas, USA; www.plusonerobotics.com)  takes a novel approach to solving this issue by incorporating a complete vision system with a robot, AI-based software, and remote supervisor software that loops in a human “Crew Chief" for 24/7 support.

In a busy warehouse or facility where rapid pace is required to keep up with demands, high-speed parcel induction systems, such as a solution from Plus One Robotics, help optimize picking and placing of mixed parcels, bags, and products. Mixed packages enter a facility on a conveyor belt that moves the items to a robot cell, where an overhead vision system identifies items and determines which products should be picked next. Once this information is acquired by the system, software commands the robot to pick and place the item onto the downstream location. In addition, a quality check is performed to ensure pick success.

While the system from Pick One is built with generic interfaces to remain hardware agnostic, it requires a 2D color image and a 3D point cloud for operation. Typically, this involves using Intel RealSense 3D RGB-D cameras (or 3D depth cameras) such as the 415 or 455 models from Intel, which are based on a structured light technique to capture depth measurement (D) while also obtaining a color image (RGB), then combining the data together, pixel-to-pixel, to create RGB-D images.

Plus One Robotics’ parcel induction system typically uses a 25 lb.-payload-capacity robot from companies such as ABB, Fanuc, or Yaskawa, while the company’s mixed depalletizing solution generally uses a 100 lb.-payload-capacity robot from one of the same vendors. Each of those applications also requires an industrial PC based on Intel i7 architecture augmented by an NVIDIA graphics processing unit (GPU), which can be sourced from several vendors depending on the application need. For lighting, the company uses products from Smart Vision Lights, according to Shaun Edwards, CTO and co-founder at Plus One Robotics.

Real-Time Human-Robot Collaboration

To help the system better recognize mixed objects on a conveyor, the company’s PickOne software uses AI tools to extract the outlines of individual objects within that scene, according to Edwards.

“No AI would be needed here if all the items a facility was handling were the same, but these businesses are dealing with basically anything and everything that is shipped in the world today,” he says. “AI helps solve this complex task by identifying individual items and outlining them for the robot to make a high-speed pick based on a sufficiently large training set.”

Since the company has more than 700 million picks in production, an industry leading metric, it has amassed in-depth data and training resources. According to Edwards, that has allowed Plus One Robotics to train its AI model over time with significant amounts of data. It does not rely solely on AI, however. The robot system is also connected to the cloud and remotely monitored by a crew chief 24 hours a day, 7 days a week, who can intervene when AI can’t solve a particularly difficult problem.

“AI is not going to solve every problem out there, so when the software can’t solve a problem, it loops in a human by sending a real-time request that typically results in a response in just seconds, which will help the robot determine its next action,” Edwards says.

Powering up to 1,000 Parcels Per Hour

Deploying advanced automation technologies such as the system from Plus One Robotics enables higher throughput at 24/7 fulfillment operations; it also decreases operational costs associated with picking by up to 70%, while allowing human employees to add value in other areas that are less repetitive and safer, according to Plus One Robotics. 

FedEx Corporation experienced that first-hand after it installed four robots to automate its small package sorting facility at its Memphis, Tennessee headquarters in March 2020, at the beginning of the COVID-19 pandemic. In addition to integrating the system into FedEx’s existing process, Yaskawa Motoman supplied the robots, arms, and grippers, while Plus One Robotics supplied the 3D cameras, software, and industrial computer.

By installing the robots, FedEx was able to keep pace with the e-commerce surge by maintaining warehouse throughput and mitigating labor shortage issues. The robots continue to improve efficiency while also freeing former package-sorting personnel to focus on less physically demanding and higher value-added tasks.

In another case, after a large apparel company found itself wasting time and decreasing throughput when handling certain package exceptions, it sought an automated solution that could increase picks-per-hour and effectively eliminate lost time. Prior to installing a new system, the retailer had people shutting down its robots to manually correct package mishaps, resulting in increased lost time due to repeated starts and stops. That’s when the company enlisted Plus One Robotics to install its system to improve the parcel induction process.

In addition to the PickOne system, the retailer leveraged Plus One Robotics’ Yonder, software that facilitates access to remote crew chiefs, to quickly resolve any exceptions and minimize downtime. After the new technology was installed, the company achieved its goal of increasing pick rates to 1,000 parcels/hour. In addition, since the system’s AI software learns continuously and the robot becomes more efficient over time, it was able to handle exceptions even faster within the subsequent six months. In yet another benefit, Yonder delivered valuable performance metrics indicating needed maintenance or training opportunities that further optimized throughput and operations over time.

More Robot Sales

Robot adoption continues to grow worldwide as companies protect themselves against labor shortages and strive to increase overall throughput, efficiency, and revenue. In fact, Association for Advancing Automation (A3) figures show that 2022 saw record robot sales in North America – to the tune of 44,196 robots valued at $2.38 billion – which represents increases of 11% and 18%, respectively, over the previous record highs in 2021. And remember: Robots aren’t taking away jobs; instead, they are freeing up humans and opening the door to better careers, according to Edwards.

“It isn’t necessarily the fact that people don’t want to do these jobs anymore, it’s the fact that many people didn’t want to do these jobs in the first place,” he said. “Robots work and people rule, we like to say. Managing a team of robots or operating one on the plant floor is a much better job than sorting packages non-stop.”