2. October 2018
Robotics, what exactly does that mean? When does a machine become a robot? And what is the difference between conventional robotics and Advanced Robotics?
The term robot comes from the Slavic word “robota”, meaning labor or drudgery. As a term for humanoid machines this term first appeared in the play Rossums Universal Robots, by the Czech writer Karel Čapek (1890–1938). Isaac Asimov, a science fiction author, first spoke of robots in his short story Runaround in 1942. The desire and idea to build automated machines, however, goes back much further, to antiquity. Leonardo Da Vinci made sketches and plans for the construction of robots. People have always been fascinated with creating machines that can perform tasks or interact with them. And there is another point of contention in the terminology: are robotics and automation synonymous? According to Thomas Christaller, robots are “sensorimotor machines for expanding the human capacity to act. They consist of mechatronic components, sensors and computer-based control functions. The complexity of a robot differs significantly from other machines in the greater number of degrees of freedom and the variety and scope of its behavioral forms”.
The big difference: the ability to act autonomously
So the biggest difference between a simple machine and a robot is the ability to act autonomously in certain situations. Looking more closely, many conventional robots today are, in the strictest sense, only very powerful machines. An example of this is the conventional industrial robot. Modern productions are impossible to operate without them. Cars are almost exclusively built by giant industrial robots, which perform the same movements and the same work steps extremely precisely and quickly over and over again. Although these robots are state of the art, they fall under the umbrella of conventional robotics. Their actions are completely predetermined, meaning that they are deterministic and not autonomous. Another example would be a washing machine. Once the button is pressed, it begins its standard operating routine. There are no deviations. The robotics of the future looks different. In a nutshell, one could say that tomorrow’s robotics is no longer deterministic, it is flexible, intelligent and autonomous. A real quantum leap. In order to understand the extent of this change, we take another step back and look at a (still) imaginary washing machine: using various sensors, the washing machine recognizes which laundry it should wash today. For example, black sportswear. Knowing this, it selects the appropriate detergent and temperature. It uses the water level of the machine to calculate the duration of the wash cycle. And if the detergent supply is running low, it sends an order to a predefined online retailer. This behaviour is a lot cleverer than a conventional industrial robot will ever be – but it’s a far cry from “advanced” because the washing machine is not autonomous. Its behaviour space is still extremely limited, lacking the ability to react to novel situations. It has been well programmed, its sensors work, but only in a specific domain. If “black” and “sportswear” do not appear in the program code, it is unable to operate.
Advanced: a Magazino robot must make its own decisions
So let’s take a look at Magazino robots and their abilities. In a warehouse, these robots can not only take over individual steps, but actually carry out all operations that a human employee would in the same position, meaning they can complete the entire picking process. Connected to the warehouse management system, they receive their pick orders via WiFi. They then navigate autonomously to the correct shelf, identify the target object with their cameras and sensors, grasp it, store it in their backpack and transport it for further processing. Paths through the warehouse and grasping motions are not predefined. In each situation, the robot has to decide for itself the optimal route through the warehouse and where best to grasp the target object. It is also confronted with obstacles and unexpected difficulties. In any environment where people work, order tends toward chaos. In other words, the environment is dynamic and subject to change. Humans are not robots, their movements and grasping motions are not computed in the same way. Humans don’t place objects with millimeter precision, aligned perfectly with the shelf. A Magazino robot must be able to work with this human chaos – efficiently and trouble-free. And having overcome precisely this challenge makes Magazino robots “advanced” in the truest sense of the word. Magazino robots can not only adapt individual parameters of their actions, but also decide to carry out completely different actions depending on the situation. The ability to adapt allows Magazino robots to deal with complex real-world problems.
An example: the Magazino robot TORU drives along a corridor in a warehouse and wants to turn left, only to find the passage is blocked by a person. TORU doesn’t just stop, it immediately calculates an alternate route to its destination. Upon arrival, it does not find the target package at its expected location. With its cameras, it now searches the surroundings to find the misplaced object, grasping it successfully.
ACROS – more intelligent, independent and effective robots
Job done, next one. This intelligent behaviour is based on the idea of not pre-programming everything, but rather to give the robot its own perception through cameras and sensors, and the ability to process this data in order to make decisions. The basis for this is the “Advanced Cooperative Robot Operating System” or ACROS for short: the operating system for our perception-driven robots. ACROS is our framework that will make the development of intelligent, perception-driven robots much easier and more efficient in the future. The knowledge that our robots collect during operation makes them smarter in real time, because they work cloud-based and share their knowledge. Our goal is to make our robots more intelligent, more independent and thus more effective. For example, with a global database of objects and corresponding grasping strategies, each robot could access this data whenever it is confronted with an object it does not yet know. This is just one example of many scenarios. Perception-driven control, machine learning and cloud-connectivity have great potential in robotics. Every day we work on further developing our robots and their capabilities. And every day we work on the future of Advanced Robotics – pushing the boundaries is what we do.