Mujin: Japanese Robotics Startup Tackling Old-School Manufacturing

Mujin Logo

A few of my robotics friends (Issei Takino, Huan Liu, and Rosen Diankov of OpenRave fame) quietly started a company in Japan called Mujin that is modernizing old-school manufacturing. Mujin uses modern software and interactive motion controllers to help large manufacturers update their production lines (using decades-old robots) in drastically shorter times compared to what would be required using crummy old motion-jogging panels. By most accounts, this is boring non-flashy robotics work. But that's precisely why it's awesome.  All too often it seems like roboticists are fixated on the glitz and glam of building robots, without solving real-world problems. Mujin is solving real, dull robotics problems in a big market with lots of money. I had the opportunity to grill them on their company and the differences between doing business in the US versus Japan -- see below for a Q&A.

I said that Issei, Huan, and Rosen "quietly" started this company, but truth be told: Mujin is already a few years old. They also raised a bit of VC funds (and are raising more!), and have already helped a "a large camera producer" improve their 60-robot production line using just 1-2 months of design time versus the 12 months it would typically take.  

They have a number of products (simulator, scanner, unstructured pick-and-place, etc). But if I were to summarize the overall vision of their product suite, I'd explain it thusly:

Say you have a production line using 20 year-old robots.  You want to improve the efficiency of the line, or repurpose the robots to a new line, but you don't want to drop a lot of capital on a new flock of robots or spend a few man-years to program 'em.  Here's how Mujin can help you modernize your setup:

1. Go in and scan the old (or new) setup using depth cameras to build accurate 3D models of the workspace and the robot, including its kinematics.  Here's a video:


2. Use Mujin's simulator to design new robot behaviors quickly in software.  Another video:


3. Download the task-based code into a new control panel. Tweak the design on-the-spot, complete with integrated simulation. Crappy motion-jogging panel interfaces are a thing of the past!

Mujin control panel


Frankly, I don't know how far along this vision they are (yet!); their solutions might still be relatively one-offs, and you can tell that some of the systems (in the videos) could use some refinement. But more importantly, I understand and agree with their vision -- they're on the right path! I'd invest if I had the resources. ;-) 

Anyway, on to the Q&A....


Travis: I clearly see Mujin's value proposition, but it's a rather "dull" product compared to the flashy robotics demos that capture journalists' eyes.  In some ways that's good (it's one of the 3 "D's" of robotics: dull, dirty, and dangerous), but I could also imagine that it makes life difficult (eg. sales and PR).  Thoughts?

Issei, Huan, & Rosen: It is indeed challenging to explain to non-robotics folks why industrial robotics is exciting and Mujin's technologies offer new possibilities for manufacturing. Fortunately, anyone who has used a robot knows how difficult it is to have a robot system run 24/7, and it's easy to show customers how Mujin's automation technologies can reduce cost. We do very focused sales by talking to factory owners and system integrators directly, so there's no shortage of interest and customers.


Travis: One of the things that fascinates me about Mujin is its focus on the Japanese market.  I'm not sure how familiar you are with US (and European) markets.... but do you have a sense for how the Japanese market (and culture) for robotics differs from the US?

Issei, Huan, & Rosen: Being in Japan gives us huge competitive advantages. Japan has the most advanced manufacturing companies in the world and one of the biggest markets for industrial robotics. More than half of the industrial robots working in the world are Japanese made, and Japanese roboticists have decades of experience using industrial robots. They fully understand the limits of the current automation technologies due to the lack of advanced software, and are willing to spend a lot of money for testing new technologies to help create better and cheaper factory automation solutions. Even though Japanese manufacturers have high standards, they've been very helpful to test and iterate Mujin's products.
So why Japan? We believe that in order to succeed in manufacturing and gain trust from the customers, we need to be very close to the end-users as they require immediate response 24/7. Customer demand drives technology progress. We also get to eat the best sushi!
As for the US... it is getting more expensive and more difficult to manufacture in the USA, and few companies can understand that automation products can materially add to a company's growth and success. This means expectations for automated systems are much higher in the USA and the market is tougher. Therefore, we believe the Japanese market and culture is the best place to nurture Mujin's products and help us prepare for the global market.


Travis: For example, what is the funding climate like for robotics in Japan?

Issei, Huan, & Rosen: Compared to Silicon Valley, getting funding in Japan is much tougher. Japanese investors are still in its early generations of venture funding and a bit more conservative. Fortunately, some Japanese investors have a lot of appreciation for factory automation and deep understanding of the manufacturing industry.


Travis: I read one of your case studies about a "large camera manufacturer."  They have an entire support staff for robot automation.  Are these types of companies receptive to change?

Issei, Huan, & Rosen: The manufacturing world moves very slowly and is reluctant to any change. Fortunately, the Japanese manufacturers realize they need automation to survive, so there's a huge urge to move forward with new technologies. They have reached the limits of "classic" automation; if nothing changes, the existing labor-intensive way of deploying robotic systems will not keep them ahead of competitors that are pursuing more-advanced robot automation.


Travis:  How about smaller companies that don't have dedicated automation engineers?

Issei, Huan, & Rosen: They represent a new market that we are trying very hard to capture. What keeps small to medium sized factories from automating is not the hardware cost; it's the system integration cost. With Mujin's technology, we make factory automation applications more affordable and more useful than ever. The majority of the factory automation cost comes from the software-related system integration process. Industrial robots are already really cheap and extremely robust. A small-size 6-joint Japanese robot arm costs less than $20k, provides micrometer accuracy and lasts for 20-30 years. However, every time the user wants to use the robot for a new task, they need to program the robot. Traditional robot programming involves moving the joints by hand to record its motion. It is very labor intensive, requires special skills, and the results vary based on individual engineers' experience. The old process is not only time consuming, but also inefficient. Mujin's motion planning technology frees robot users from tedious robot programming, and allows complex robot motions and factory layout optimizations that are impossible with manual programming.


Travis: Mujin seems to have a number of products / expertise.  Do you see yourselves focusing on one key product, or is it really an "entire system" requirement?

Issei, Huan, & Rosen: All of our products share the same core platform: Mujin Controller. We scale our business by creating highly demanded industrial applications based on the Mujin Controller. Eventually we hope that anyone can create their own custom factory automation system using Mujin Controller. We currently provide a RESTful Web API and clients in C++, python, and C# to integrate it with other industrial systems like PLCs and machine vision.


Travis: If I recall correctly, some of the robots you work with are older than me! Can you share some pictures (or videos) of these old robots before and after the Mujin upgrade?  I'd be really curious to hear about how much better they are now.

Issei, Huan, & Rosen: One of our applications is a motion-planning-enabled control system for multi-axis machines, such as laser cutting systems. (See the photo up above.) The old controller was made in the 1980's, and users needed to program the multi-axis laser cutter on a display that shows at most 252 characters. Our new controller has a touch screen pendant with 3D visualization of the laser cutter and its goal trajectory. The user can program the laser cutter with our high level Industrial Task Language (ITL) specifying only the way points and the types of the curves, we generate the actual trajectory with high accuracy and collision avoidance.


Travis: Any thoughts on the Google robotics acquisitions: Good, bad, undecided?

Issei, Huan, & Rosen: It means that it's an exciting time to do robotics! Hopefully more talented engineers will start working on robotics instead of making mobile apps or building websites, and more investors will invest in something that will really make the world a better place.


Travis: Can you think of any other big robotics opportunities (outside Mujin's area of expertise) that are ripe for robotics disruption?

Issei, Huan, & Rosen: Retail is an interesting area where robots could be used to demonstrate and sell products; moving robots really attract a crowd! Businesses related to e-commerce are also good places to apply robotics technologies; there is abundant funding and other opportunities for automation.


Travis: Mujin seems to be going after more-traditional manufacturing markets. Recently, with Rethink robotics and Redwood robotics, there seems to be some push toward "flexible" (smaller-scale) manufacturing. Why is that?  What are your thoughts about this space (and compared to more-traditional manufacturing)?

Issei, Huan, & Rosen: Reusing existing hardware will make customers more willing to try our technology, and help iterate our technology faster in the market. We can also take full advantage of existing integration support. In addition, we think that there are a lot to be done to help improve the "legacy" industry before going to create a completely new market. There is plenty of low hanging fruits in the traditional manufacturing world! The tasks may not seem exciting at first, but there are a lot of engineering challenges and lessons to be learned that we need to overcome before we can aim for flexible manufacturing. The Mujin team wants the industry to enjoy the latest progress from robotics research, and make the scholars proud by demonstrating which cutting edge technology works 24/7 and can contribute to the world's manufacturing productivity.