Mujin Inc.: Revolutionizing robotics and automation

If we could start with a quick introduction to Mujin. What is Mujin's core business and what are some of the advantages that have allowed your company to experience the rapid growth you have since establishment?

We already opened a US office in 2021 as well as a European office last year. In America, our business is booming right now with direct customers. including one of the biggest retailing companies between America and China, and right now as you might know, tensions are running high between these two superpowers. Although Japan doesn’t necessarily have a strong attitude, we are in a position to sell to both countries.

More than half of our company is non-Japanese, although our headquarters are based in Tokyo. There are many foreigners working there including people from America, Australia, and Europe. Even our Chief Technical Officer (CTO) and Co-Founder, Rosen Diankov, is an American genius, and he has brought his American style of management to this company. I personally have experience with the Japanese manufacturing management style. Synchronizing these two styles is usually very difficult, but still, we have managed to combine the two very well.

Company culture is another aspect that is very important, and if you had this type of startup in Silicon Valley you might expect the engineers to act a little spoiled. Robotics is a serious business for us, and we aren’t playing games. We have to deal with robots that are in place in real factories or warehouses, and if something goes wrong there could be significant damage, not only to equipment but also people too. Monetary damage is huge for these major customers and this might be why the barrier to entry in this industry is so high. Whether you are from MIT or Stamford, the company culture dictates that on-site visits need to be made. This whole culture of Mujin is crucial to our operations, and this is because our CTO doesn’t hesitate. Without this kind of culture, you can hire all the smartest people in the world and you won’t achieve anything because they won’t want to solve real problems.

Right now, Mujin is mixing these work cultures with the mass production that we are doing in China. I like to refer to this as a triangle, and right now it is working very well for us. Usually in the robotics field, the people who produce the robots are also the platform makers and therefore are referred to as robotics makers. Back in the day, everyone had their own operating systems, such as HP and IBM. Mujin is the only company that has been able to unite third parties together under one operation platform. We would say that we really are the best in the world, however, I understand your doubts if you have them because almost every company in the world claims they are the best. The proof, however, comes from deployment, and we already have thousands of our robots doing real mass production in the real world. I think when you think of intelligent robots in actual deployment, we have the most in the world. By the way, when I say intelligent robots I mean robots with sensors that can make decisions.

We have a very interesting business model and exceptional international teams. In fact, currently, we have 340 people across four countries working for Mujin; those countries are America, Japan, Europe, and China. We currently have more than 550 different patents. I would say that our patent policy is quite interesting because we apply for patents in America first. That is because in Japan businesses don’t tend to sue each other, but America is almost the complete opposite. The approach is to obtain the patent in America first before copying the applications for our other locations. Most of those patents actually relate to software rather than hardware. Although Mujin provides both hardware and software, our software is the most original aspect of our business. It is a similar approach to Apple. When you think about it, Apple doesn’t have its own factory but they have their own products and software. They design the products and then have an OEM produce before the original software is applied. Our main product is the brain part of industrial robots, often called a controller. We also produce the eye parts, or the sensors as well as the hand parts, also known as the rig. Everything is controlled by Mujin’s original, full-stack software, which includes full front-end and back-end development. Since our solutions are very strong, customers choose Mujin again and again.

We founded Mujin back in 2011, not long removed from the Lehman Shock, and since then we have grown from strength to strength. In fact, in November 2023 we just completed a strategy partnership agreement with AEON, with Mujin being the partner in that regard. Japan Post joining us as well as a very big Japanese retail company. As you can see, things are very exciting right now and we are growing very fast.

As the CEO I’m actually in the minority here being Japanese, and all of our board meetings are held in English. We like to think that we are gathering all the smartest people from all over the world and uniting them for a common goal. Those people are not isolated at Mujin because we all speak English. We are also in a unique position because as a startup we’ve retained more than of our executives, a rare achievement in this day and age. All of our executives have won various robotics and software competitions around the world, major achievements in our fields. Everyone is a genius. In fact, Mr. James Kuffner, COO of the Toyota Research Institute is one of our advisors.

Basically what we are doing at Mujin is shifting hardware-centric industries to software-centric industries. We are taking the same approach in terms of robotics. Software and solutions are more important than hardware.  

One aspect we found interesting during our research is your Mujin 3D Vision System, a core part of your control system, a proprietary technology that utilizes numerous sensors. Achieving a high accuracy using this technology must be very difficult, so how was Mujin able to overcome a myriad of challenges to ensure the high accuracy of the Mujin 3D Vision System?

The vision system itself isn’t actually a big part of the automation. With human beings, if we can see something we can pick it up, and that is because there are so many sensors throughout our bodies that allow us to do this, it is just that we are unaware of those sensors. For example, our hands will feel an object and send information to our brains about the size, shape, texture, and weight of an object. That is then coordinated with visual information sent from our eyes, as well as many other factors. Seeing something is actually just 10% of the senses. Remember, when you pick something up, your body is coordinating muscles in your hands, arms, legs, waist, and heart all together in unison, and these muscles are adjusted based on that initial prognosis from your senses. This is done in a split second without you even really being conscious of what is happening inside your body. If the object is suddenly heavier than imagined, your body adjusts on the fly to that. The same could be said for our robotics, but compared to the human body we aren’t nearly as accurate. That is why the sensors allow for so many adjustments. The amount of sensors in our robotics will amaze you, there are so many. This is what I mean by saying that the vision sensors are only a small part of it.

Just like the human brain receives information from all over the body, so too does the Mujin Controller. That information has to be processed and instructions then have to be fed to the various parts of the robot. Controlling, processing, analyzing, and instructing this much data is a huge task, and that is why creating intelligent robots is such a monumental task.

There are so many people working in factories and warehouses conducting very dangerous work, so our mission is to replace them, freeing those workers to conduct more human-centric work. In today’s modern world, relying on human labor is reaching its limit. Companies also can’t keep up with wage inflation, claiming that a single worker in a warehouse costs them USD 33 per hour. They cannot bear that cost so they would rather automate.

I think you also have to consider human error. In the fast-moving times of 2023, one small human error can cost major companies millions in time lost. Imagine the cost to a car maker for example if they had to recall a particular model of theirs because of a factory defect resulting from human error. Additionally, on the theme of EVs, there are so many car makers right now, big and small. Before equipment makers only tended to deal with one of the big 5 major makers, but with so many small companies producing vehicles they have to deal with so many customers. This means you cannot do mass production either. If you ask one person to make different kinds of products what do you think is going to happen? That person is going to get overwhelmed with the different products and mistakes will happen, which is almost inevitable. This is going to be a problem that persists for the next 20,30, or even 50 years, and that is because it is a social problem.

Robots have existed for over 50 years now, so I often wonder why people are still working in these manual labor positions. The robots are simple, and what you program them to do they will repeat. I think the key to this question is intelligent robots because while some of these factories' tasks are repetitive, they still require decision-making on the fly. These kinds of robots are flexible and can adapt to their environment.

Every single customer I’ve ever spoken to has expressed the desire for intelligent automation, but I think the bottleneck that has held the industry back for many years now is the software side. Operation systems usability has historically been challenging, often requiring users to learn special languages. Even simple operations often require a programming background from engineers, something our robots do away with. In some of our applications, a robot might have around 30 sensors, all sending information and signals to the controller. Feedback and logic loops are all things that require some sort of sensor information and decision-making, all coming from deep robotics technology and knowledge.

Next, you have elements like error recovery, and as you can probably guess, there are so many kinds of errors. You have to pre-program the robots on how to recover from those errors. This requires significant foresight into the kinds of errors that might occur. Obviously, this comes with experience, but still, it is impossible to remember every type of error. The good hardware is already there, so with error recovery you really need good software integration in order to have it become effective without human input. Think of it like a computer without Windows. Without Windows you can still use the computer but only very simple applications. In order to do something more complicated you need an operating system (OS), or a front-end to bridge the gap between users and machine code. The user interface (UI) is like a mouse or a keyboard. I think this kind of intelligent operating system like Windows is the key component that is missing from our industry, so that is what we are making. In fact, this idea of bridging the gap is the initial catalytic idea behind Mujin.

As you might be very well aware at this point, Japanese companies struggle with software. They are known to be very strong with hardware, think back to devices such as the or vehicles from giants such as. The big weakness however comes from the software side that accompanies these terrific hardware devices. To put it frankly, historically Japanese companies have lagged within software development. Even big industrial firms like Caterpillar use overseas software because the Japanese software simply isn’t good enough. Literally, the only software I can think of that Japanese firms have had any success in is known as Numeric Control (NC), software for machinery. Perhaps this is why so many computers use either Windows, Linux, or Apple OS, and mobile phones use either IOS or Android. All of these operating systems are American in origin. I don’t think a Japanese mobile phone OS even exists. Interestingly, when we use a third-party robot we don’t use their software at all.

When you see our 3D Vision competitors in America and China you will see that they are only making applications. They are moving their applications onto someone’s platform, whereas we make our own platform, completely from scratch. We don’t need the robot makers platform, and it means that we are in a very unique position when negotiating with robot makers. Think of it in a similar way to when you select a phone. It is either IOS or Android. When you buy an iPhone, you don’t get to select the battery maker, it is what it is.

|

|

Previous Next

Just to clarify, are you selling the whole solution including the hardware and software? Is the majority of your sales or would that be just software-based solutions?

A majority of our business is for both. Our software is combined with robotics produced with our branding in order to provide customers with a holistic solution. The software side is completely developed in-house here at Mujin.

One thing we are focusing efforts on right now is teaching the robots pre-programmed motions, allowing for much smoother motion from the robots. The combination of six motors and our software essentially gives the robots constraints, which will then over time allow the robots to teach themselves this smooth motion. This is what I mean by pre-programming motions. The computer is always thinking about the best route to move an object, adjusting its motors and sensors to assess the situation. What we do here is plan the motions, but a robot maker teaches the motions. This approach is very different. I think our approach really demonstrates how some of the motions of the robots are impossible for a human to teach. Using the physics model as the only constraint, the software allows for robots to intuitively discover motions that are not taught, and are physically impossible for humans to replicate.  

Without the platform that we offer, our competitors can’t keep up with the number of applications we can apply. Their robotics cannot learn motions automatically, and just by seeing the motions side-by-side, you can see the difference.

Mujin started with logistics and within that industry, there are so many stock-keeping units (SKU) to account for. We, in fact, still have systems in the industry that we put in place in 2017, and those are very fast and have yet to require any form of reprogramming. This is where the 3D vision system comes into play. The robot is responsible for figuring out how to handle all of these different SKUs using the methods I’ve mentioned earlier. In fact, I mentioned error recovery earlier too, and we have a stock of over 10,000 error recovery codes right now, a number that wouldn’t be possible with just human beings.

I think a huge key point of the robotics industry at large that is gaining prevalence right now is the simple fact that we’ve moved beyond human experiences to guide our automation. We are now using integrated experiences developed by both the hardware and software. This honestly allows for better stability in our automation solutions, and with the number of combinations of advanced software with robust hardware, I believe that Mujin has created a recipe for success.

Basically what you are suggesting is that because of your robust software and hardware, nobody outside of Mujin needs to worry about troubleshooting. Is that a correct assumption?

Yes, we minimize that, so that would be true as long as the hardware doesn’t break. Most of the time we handle things remotely using our platform.

From a customer’s point of view, if they were coming to you looking for a new automation solution, would you recommend a whole suite of both hardware and software?

Yes, I would recommend both hardware and software. There are two customer layers that we consider. Firstly you have end-users and then you have consulting teams. Literally what we are offering here at Mujin is a full stack solution, from design, production, software, integration, and finally after support. A big difference however is the size of the company. If it is a big company we will tend to go directly ourselves to conduct these services, however, if it is small we tend to use affiliated system integrators for the workload. The key is that all the inquiries come to Mujin first, and choose if we do things directly or if we use a system integrator. 

It is interesting the perception of robots that differs in Japan and the US. In the US, because of Hollywood, there is a misconception that robots are going to one day decide to kill us all and take over. It is probably because of the popularity of movie series such as Terminator there is actually a fear of robots in the West. Japan on the other hand has a real affinity for robots because of anime such as Doreamon and Gundam. Japanese people see robots as helpers in our daily lives. I think you have to remember that a robot is only part of an automation line. It is not designed to eliminate human life, therefore it will never do that. Also, people tend to fear the idea of losing their jobs to robots too, but this sort of simplification of tasks has happened time and time again throughout history. Before word processing software there were typewriters and countless women trained in typing. These new technologies are designed to simplify tasks and allow humans to dedicate more time to pressing matters. Productive and creative jobs can be followed through because repetitive jobs can be handled by automation.

Automation has been designed throughout history to improve human lives and just part of that process are robots. If you look at something hard enough, chances are it can be considered hard enough. Things like pencils, cars, TVs, and computers. Think back in the past they used horses and carriages and then came the steam train, which allowed people to travel much further at faster speeds. The horse riders union opposed the usage of trains, but then in the US in particular, the number of people working in the train industry increased. The same happened when cars came to prominence. Those pieces of history are essentially automation and were responsible for saving people so much time.

In Japan right now there is a huge deficit in human labor with the low birthrates and aging population. It is a massive issue for the Japanese government and companies across the country. The fear of automation doesn’t exist there simply because people understand that we aren’t replacing humans, instead, we are compensating for the lack of humans.

As a robotics solution provider, what are your thoughts on potential regulations and laws regarding robots? What benefits or concerns would these regulations present to Mujin?

I think the emergence of simple service robots is something we are already seeing in restaurants or convenience stores, but more complicated, android-like robots are going to come much further into the future. This isn’t because the technology is dangerous, rather it is because the technology hasn’t developed enough yet. Factories and warehouses are controlled environments, but a service robot would need to work in an uncontrolled environment. Right now there are things we can’t do even in a controlled environment, so trying things in an uncontrolled environment is out of the question.

What your company is providing is clearly applicable to so many different sectors. Where do you see the most potential currently?

Our aim remains the same; factories and warehouses. Our problems in those markets are very clear to us right now, meaning that we can set clear goals and standards for Mujin to move forward within the logistics sector. Of course, customers know what they want, but at the same time, we understand what we can supply. In the service industry, customers struggle to define exactly what they want in terms of automation. 

You mentioned earlier about working with Japan Post, a business that is traditionally very analog, and you are going into that environment and totally revamping their operations. Are you proposing to them that they should automate or is the customer telling you what they want?

This will depend on what they desire. A big part of their budget may be allocated to human complexity, so in that case, we wouldn’t propose an entire solution. The services provided will depend entirely on the customer.

One particular product that I wanted to showcase was our truck unloading solution. As I mentioned earlier, we have a common platform where we can bring a lot of applications to. If we can buy the hardware from amazing robot makers or suppliers we will, but if we cannot find it we need to make it ourselves. This is actually something we made. We call it the TruckBot., and everything is handled by our one, unifying platform. With the cameras and sensors on the hardware, on the platform, we can monitor issues too for issues. Literally, as soon as a problem occurs with a client we can troubleshoot remotely first.

Moving forward, what other regions beyond America are you looking to further expand?

Right now, it is definitely America, and this is a very Japanese answer. Japanese firms follow the carmakers. Aside from that, Europe will definitely be a target in the near future.

What is your strategy in order to take full advantage of these markets? Will you look to establish joint ventures or partnerships in order to unlock the full potential of the US and Europe?

As an organization, we don’t push the Japanese style. Local people manage our local operations. Another strategy is partnerships, and with automation, it is really important to acquire big customers. One big customer could unlock so many potential markets.

Imagine that we came back on the very last day of your presidency and had this interview all over again. What goals do you hope to achieve by the time you are ready to hand the baton onto the next generation of Mujin executives?

Of course, I would say the ambition of everyone here at Mujin is to see everyone using Mujin systems as their automation solutions. The ultimate dream is to have everyone pick from a Mujin catalog, basically, pick the hardware, and it works automatically because of the Mujin platform. Essentially what we are trying to achieve is a completely unified ecosystem that can cater to any automation need that any human has ever dreamed of. That is the ultimate dream. We want people to talk about Mujin in the same conservations as Apple and Amazon.