When I first visited the lab of Toshio Yanagida at Osaka University in 2010, I was very surprised by how the researchers there managed to manipulate single proteins utilizing laser traps. I was almost equally surprised as a very cute and human-inspired robot was calmly waiting in their conference room to greet me. However, this chance encounter, with the only exception of my parent’s robotic lawnmower, has been the full extent of my hands-on experience with robots until recently.

In the meantime, however, I must admit that I might have developed a serious crush on the topic of robotics. While the use of robots swaps over from mass production employing specialized machines to lower-volume manufacturing, an interesting trend emerges — robots start to become more flexible and at times they have some sort of machine learning integrated to perform simple tasks. Slowly, but surely the machines become more intelligent. Naturally, as a seasoned machine-learning (ex-)professional, I am more than excited about the promise that lies in the intersection of robotics and AI. But before we discover the interesting world of artificially intelligent robots, let’s first build up a basis and understand the ways of the industrial robot industry.

Robots have become an integral part of manufacturing since the first robotic company opened its door in the 1950s. They are one of the key reasons behind the ever-growing quality and productivity we have experienced in the last decades. In 2022 the International Federation of Robotics (IFR) expects a staggering number of about 3.8M operational industrial robots worldwide.

Who can spot a trend in Fig. 1? The growth of annual shipments of industrial robots starts to pick up considerable speed after 2010. Statistics by the IFR show that this is to a large extent due to the technological catchup in China where the number of robots grows at an impressive rate of 30% per year. There, already 27% of all operational industrial robots could be found in 2018.

The largest, but by far not the only manufacturers of industrial robots are ABB, Omron, Fanuc, Kawasaki Robotics, KUKA and Yaskawa. In Fig. 2, we can see the revenue of these robotics giants. Their main customers are still the highly-automated electronics, automotive and metal industries.

In the last years, Chinese producers are demanding their piece of the cake. While their market share is still small, manufacturers of industrial robots like Honyen, Siasun or Estun are slowly gaining market share in their home market. To get a feeling of the proportion, however, let me mention that in the first three quarters of 2019, Fanuc alone sold more than 3.5 times the number of robots in China than all of its Chinese competitors. As industrial robots have life spans of decades, the proven long-term reliability and robustness of the incumbents seem still to be very convincing.

There are different type of robots like SCARA, delta or cartesian coordinate robots. An intuitive example is a so-called articulated robot (see Fig. 3). It is modelled after a human arm and with its six degrees of freedom, this industrial robot is very flexible and therefore amongst the most popular robots for industrial applications. Robots like this are for example used for welding, assembling, sealing, material handling, picking, cutting, painting or spraying. We probably have all seen this type of robots in videos of the impressively automated production lines of car companies or similarly advanced industries.

The robot’s design consists of five main components: controller, sensors, manipulator or robot arm, end effector and drive.

The controller can be seen as the brain of the robot. It can be connected to a computer and serves as the interface between the person that sets up the system and the robot itself.

The robot perceives its surrounding via sensors. These may be microphones, cameras or pressure sensors. Feedback of the sensors may be used by an industrial robot to perform its programmed task.

A robotic arm is used to position its end effector. In our example of the articulated robot, the arm is modelled of the human arm and mimics shoulder, elbow and wrist. It has six degrees of freedom and therefore exhibits a wide range of possible motion.

The end effector acts like the robot’s hand and is located at the end of the arm. Depending on the purpose of the robot, the effector may be a welding torch or a gripper.

The robotic drive is the motor that moves the robot parts. Common drives are powered hydraulically, electrically or pneumatically.

Reading this post, you may think: “Nice! Where can I buy a robot to automate my every task?”. But be aware! This industry is very advanced and complicated.

In an ideal world, you would be able to buy a controller from ABB, sensors and a robot arm from Yaskawa, your favourite end effector from KUKA and a drive from Siasun and combine your eclectic robots to serve your very specific needs. In the real world though, most companies use different programming languages and not all of their APIs are open. Robot manufacturers do want to be more than mere producers of robotic parts and aim to push the software side of their businesses. That is why in a production line you mostly see complete robots from different vendors working next to each other. Fortunately, there are attempts like the Robot Operating System (ROS), which despite its name is rather a middleware than an OS, aim to unify the software level. Nonetheless, it is pretty unclear at the moment how its market adoption will develop.

Companies that keep a cool head and help you with setting up your production line are the system integrators. They will come to your facility, understand your needs and make the hardware of different providers work together. Their business is to get your production line ready and they get paid like kings. If you are interested in the largest system integrators and their revenue, feel free to consult this list by Control Engineering.

There are different use cases and strategies for automated manufacturing. They all have strengths and weaknesses and involve very different costs and timelines.

Custom machines have been the enablers of mass production and, in many cases, it still makes sense for factories to design and build special-purpose machines. A good example is a factory for cigarettes which almost certainly uses specialized cigarette-making machines. Product lines in such factory change about once every two decades and the high initial investments are recovered in the long run. Because of specialized machines, very few labourers are needed to produce vast amounts of cigarettes. Such bespoke robotic systems make sense for high-throughput systems with a long lifetime.

A slightly different case is the production of cars where a new model may be produced every couple of years. In this case, production lines will need to be adapted to every new model. It is not economical to design and produce bespoke machinery for every part of every new car. Automotive companies, therefore, make uses of more flexible robots like our articulated robot depicted above. This flexibility, however, does come with a price tag. Because such robots can be re-used, setting up a new assembly line is complex. System integrators may invest months to set up and program the robots in a complex production line. The initial high investment for this has to be amortized over the years. Such a use case is characterized by standardized parts and production lifecycles in the order of magnitude of several years. Traditional robotics are designed to work autonomously with safety assured by isolation from human contact.

While mass production is an obvious case of robotic automation, currently high-mix and low-volume scenarios are mostly performed manually. In this area, the initial investment of programming and setting up robots in the traditional way may not pay off. Here is where we see lots of development: the usability of robotics software is simplified considerably and technologies like machine learning are used to teach the robot tasks that are not clearly defined. A nascent trend are the so-called cobots — collaborative robots. Currently, they represent about 3% of the robotic market, growing slowly. Cobots are intended to interact with humans and work near them. A company fielding this long tail of robotics is the Danish Universal Robots. As an example, their UR5e is shown in Fig. 4. Isn’t she a beauty?

Even only scratching the surface of the topic of industrial robots, it already becomes clear that robotics is a fascinating field. Interesting things are bound to happen if robots become more versatile and intelligent. I am especially excited to see how the flexible cobots will be used and adopted within smaller productions lines and how robots will find their way into our everyday lives.

In follow-up posts, I will dive a bit deeper into the competitive landscape, the overall market and the startups that appear to change robotics for the better.