Melonee Wise and her three colleagues had worked together for seven years when they started thinking about logistics. Meanwhile, Tech-RX started a company called Fetch Robotics. When founder Steve Hogan needed help, he called on Wise and her team. Just days after starting in August 2014, Wise and her team came on board and eventually took control of the company. Today, Wise is CEO.
“We wanted to go after a semi-structured environment,” Wise said. “With an unstructured environment, it’s hard to create rules for people to follow. In a manufacturing or logistics environment, it’s pretty cool.”
The company completed its first round of funding in February 2015 with $3 million from Shasta Ventures and O’Reilly AlphaTech Ventures. Four months later, SoftBank led a Series A round that produced $20 million for Fetch Robotics.
Mobile Robots With Intelligence
Fetch Robotics offers two services: mobile robots and data surveys.
The mobile robots are distributed throughout picking areas and transfer materials like a virtual conveyor. This results in a 20%-30% increase in man hours for their customers while taking just 48-72 hours to set up. The customer pays for the robot and the monthly cloud fee.
Robots possess a lot of intelligence for driving, navigating, and avoiding collisions. In the cloud, the company coordinates the movement of robots, scheduling, and ensuring the tasks are completed on time.
Fetch Robotic’s second product is a data survey, allowing partners to automate the collection of data. For example, a robot utilized in retail can collect pictures in stores for tracking analytics. People taking pictures of shelves is inefficient. It isn’t repeatable or reliable because of human error. Robots can do it consistently and repeatedly for hours on end.
Using Robots to Conduct Data Surveys
Data survey employs some form of computer vision, whether laser scanner or 3D time-of-flight cameras, on a small (15” x 14”) semicircular robot.
“We don’t support learning,” Wise said, “we do training with a training dataset and decide what behavior is going to be. Robot safety requires it. New policies can’t be learned without support as it could be dangerous. So they are never unsupervised.”
Fetch Robotics programs its own intelligence.
“Navigation, mapping, scheduling, and multi-robot coordination programs are all written by employees of Fetch,” Wise said. “One key difference we have over other new robotics companies is experience. Some companies in the logistics space decided to hire a robotics team whereas he have robotics expertise.”
With mapping technology, Fetch Robotics builds large maps of warehouses without human intervention, and they do it quickly.
“We don’t do hand editing,” Wise said. “We utilize on-flight 3D cameras mounted on robots. They see low-to-ground obstacles, which makes the robot safe in dynamic environments.”
Wise also said the videos you see of pristine warehouses is a fiction.
“Warehouses don’t look like that,” she said. Her robots are capable of dealing with the mess of life to fulfill their missions.
The Lowdown On Fetch Robotics Technology
Fetch spent a lot of time putting software in the cloud to manage real-time data and support, running off Amazon Web Services (AWS). Not many companies have the same capability. With 45 employees, they manage the entire company and its fleet of robots. Generallly, customers employ between five and 50 robots at a time depending on the size of the facility. Fetch Robotics, however, is capable of more.
Wise said there are a couple of ways to monitor robots. Typically, they are on a schedule and operate automatically without supervision with some checks in the system.
“We can drive around collecting data all day,” Wise said. “We have scripts to monitor, but if there’s a problem on site, customers can email us. We look to see what is happening and fix it quietly in the background. If something more serious happens, we alert the customer proactively rather than waiting for them to contact us.”
Wise gave two examples of problems that might occur with their robots. If a water pipe bursts, it could do damage to the robot. In other cases, a robot could get confused about its environment or trapped between obstacles.
“One time, two forklift guys parked on each end of an aisle and the robot ping-ponged back and forth three or four times trying to get out. It’s gratifying to know we have a way of catching that and responding without the customer needing to call us.”
With some customers, Fetch Robotics sees a 20%-30% increase in picking lines per hour. Surveying customers have seen read rates accuracy jump from 92% to 99.5%.
In manufacturing facilities or warehouses where workers must walk long distances, if the worker has to walk more than three minutes to drop a package off, that’s too long. Robots can do it faster. This frees up the human to do the job they are valuable at, like assembly. If data has to be collected daily or hourly, a robot can do it with RFID scanners and cameras. Again, the human is able to do the more complicated tasks.
Wise Looks at the Future of Robotics
“The minute you make a prediction it’s wrong,” Wise said. Nevertheless, she predicts that automation will increase in the next five or 10 years. “It’s hard to keep people employed. About 10% of logistics jobs go unfilled because nobody wants to do them. Automation is a good place to start because it increases the efficiency of people who want to do the job. If they are older, with back or knee problems, they can do their job longer and won’t have to retire early.”
Regarding Fetch Robotics, Wise said they’re working on some new initiatives.
“We need more engineers, more sales people, distributors, and integrators in several countries,” she said. “A lot of stuff goes into making the product.”
The sales cycle for the data survey product is 9-12 months. For the virtual conveyor, it’s 1-6 months. Because of the complexities of the service, it takes time to educate customers on how robots can make them more efficient.
Written with Nicki Jacoby.