Robots that move on their own must be able to adapt to the dynamic and challenging conditions in the supermarket. “My research focuses on using artificial intelligence to make machines smarter and more reliable by teaching them symbolic knowledge,” says Hernandez Corbato. “The goal is to develop robotic “brains” for intelligent robots that can be trusted to work alongside people, because they can explain their decisions.”
The supermarket is a place where unexpected things happen all the time. Not only are there thousands of products with different shapes and looks, but there are also people going in and out. How can a machine operating independently handle this safely, efficiently and intelligently? By activating symbolic knowledge that we humans also use, Hernandez says. “We recognize a tray with four legs under it as a symbol: a ‘table.’ We don’t need a picture of it. When we code this ‘code language’ and make it suitable for robots, it can perform more complex tasks.”
One look with the eyes of the camera and the robot knows that it is facing something on which plates and cups can be placed. Based on this, she can decide what to do. The acceleration that results should enable robots to perform multiple actions at the same time: move around, pick up and move objects, and ultimately communicate with people.
For Hernández, the AI research program for Retail Lab for supermarket chain Ahold Delhaize brings together all the magic about artificial intelligence. Retail requires robots to use a variety of skills: perceiving the environment, navigating around it, manipulating objects or cooperating with humans. For him, it’s all a matter of the algorithms needed to make a machine respond as intelligently as the human brain. As an autonomous robotics software specialist, he actually won the Amazon Picking Challenge in 2016 with a team from TU Delft. On this occasion, a robotic arm placed products from a container in place on one of the shelves.
“Retail requires robots to use a variety of skills: perceive the environment, move around, manipulate objects or collaborate with humans.”
– Carlos Hernandez Corbato, Researcher
The “supermarket robot” is considered more difficult. It requires moving from a static factory environment to the dynamics of the store. The traditional way, in which bots learn from the data they collect, is quite cumbersome for that. The bot will already be stuck in the inventory management. Programming a custom robotic treat for each orange, bottle, soup can, milk carton, or option would be a lot of work. “We want to inject symbolic knowledge into the bot’s operating system all at once,” Hernandez says. “If this knowledge is available, the robot can constantly adapt to its changing environment. For example, by downloading a different hand motion.”
The bot should be able to choose a different algorithm independently if it encounters a problem along the way. So he can pick up a can that falls from his hands, or change his fist slightly when picking up an unknown object. Technicians have already set up a test shop where the “Tiago” robot can practice. In about five years, it should offer a machine with a movable base, arms, and two eyes for a camera, which independently refills supermarket shelves 24 hours a day. And he must be able to do this in all conditions, day and night.
The sensor is broken
The latter is not only applicable to the supermarket robot. In fact, every robot should have a next-generation operating system to better handle changing conditions. Hernandez Corbato: “In addition to integrating different robot skills, the cognitive skills of robots need to enable them to think about these skills, understand how to use them, and what are the consequences of their actions. In short, we need to give robots (or any intelligent autonomous system) self-awareness so that we can of trusting them.”
“We need to give bots (or any built intelligent autonomous system) self-awareness so that we can trust them.”
– Carlos Hernandez Corbato, Researcher
It is the basic idea behind the European Metacontrol Project for ROS2 Systems (MROS) recently completed by the Cognitive Robotics Division. The AI technique that Hernandez used for this purpose is called the metacontrol method. Describes the characteristics and skills of a robot in an organized way, so that the robot can use the knowledge to adapt and overcome problems.
As part of this research, he developed multiple prototypes of these next-generation robots with Bosch Corporate Research, Universidad Rey Juan Carlos, Universidad Politecnica de Madrid and the University of Information Technology in Copenhagen.
Do they work better than traditional robots? “Yes, he sailed more safely and, thanks to his symbolic knowledge, was able to adapt to the conditions. When one sensor broke, he switched to another independently,” Hernandez says excitedly. “This is where we want to go: a robot with intelligence.” enough to deal with failures.”
Tags: c- research innovation
Joost van de Loo – Strategist at RoboHouse
#Shelf #Storage #Robots #Autonomous #Movement