2. October 2018
Before you know where to go, you should know where you are. This applies for something as simple as going to the supermarket, to the big decisions in life. And it also applies to our robots.
In order to move successfully from point A to point B, they must first localize themselves, then overcome obstacles and new situations along their way. To do this, we have equipped them with intelligent algorithms and collective perception.
The more freely robots intend to move, the smarter they must be. This includes their ability to navigate autonomously. In Magazino’s case, our domain is the warehouse and navigation consists primarily of two things: firstly, finding a way through the aisles while avoiding obstacles, and secondly, arriving at the right shelf. Once there, the robots have to locate the target object on the shelf and then grasp it successfully. This poses a number of challenges. For comparison: AGVs, or Automated Guided Vehicles, which support goods in warehouses and transport objects from one point to another, do not have to navigate intelligently, but are instead guided by lines, magnetic strips or other predefined tracks.
Efficiency through intelligent software
Magazino robots are different. They don’t need a complex rail system or a warehouse designed specifically for them. They can be adapted to work in any warehouse environment and their intelligent features make it easy for them to find their way around. They move freely and have the ability to navigate autonomously.
A fundamental challenge in mobile robotics is called SLAM. SLAM stands for “simultaneous localization and mapping”. This is the name given to the robotics problem in which a mobile robot must simultaneously create a map of its environment while estimating its position in this environment. To reiterate, before navigation can take place, a map of the environment must be generated, and then the position on the map estimated. Our robots do this by perceiving their surroundings with various sensors, all the while updating the environmental map. Once the robot’s position is known with sufficient certainty, a route to the destination can be computed.
The advantage of working together
This is already feasible for a single robot in a completely static environment using conventional SLAM algorithms. However, warehouses in the real world are anything but static. People work in them and change the environment continuously; goods are moved, new obstacles come and go. In order to work reliably and smoothly, even under these circumstances, the robots need one more skill: the ability to adapt. To do this, they need as much information as possible about the current situation in their environment. With this information, they can then adapt their paths and operate more efficiently.
Our approach is thus: cloud-based mapping and localization in dynamic warehouse environments. This concept is based on the fact that individual robots can collect data about their local environment during operation. Each time a robot travels through an aisle, it perceives its surroundings through sensors and cameras. If it encounters an obstacle en route, it stops and looks for an alternate way to reach its destination. However, if a robot doesn’t have to discover the obstacle for itself, it can plan its route accordingly and arrive at its destination faster. So wouldn’t it be fantastic if Robot A could share its knowledge with Robot B? Definitely. That’s why, in cooperation with Google Cartographer, we have further developed a real-time localization and mapping library. In concrete terms, this means that every robot in a warehouse sends SLAM data to a cloud and simultaneously compares its stored inner map of the warehouse with data in the cloud. The robots in the fleet can therefore share their knowledge of the environment and use it for local optimizations. One could say that the fleet perceives its environment collectively and is constantly improving together: Magazino robots are real team players, just like their developers.