A 90 cm robot with a face on a screen is being offered a place beside your child. It can tell stories and take part in English learning activities. Its maker says it can recognise family members and remember earlier conversations. The awkward product question arrives the following evening: will anyone want to talk to it again?
A Chinese carmaker has opened global pre-orders for this household companion. The automotive parent interests me because it brings experience of making a product repeatedly and supporting customers after the sale. Turning those capabilities into something a family wants to live with is a separate piece of work.
A reminder has to earn its place
The robot is designed for interaction with children and older people. Members of the household can set reminders, and the company says memories of previous conversations help it continue an interaction. That gives us a more useful product question than how well it walked during the launch.
Imagine a child asking the same question every evening. Remembering that conversation could make the machine feel useful tomorrow. Without that continuity, the family may keep meeting the same demonstration.
A robot in a family home has to earn its place after the novelty wears off.
A medicine reminder makes the challenge easy to understand. A phone already does that job. The extra machine needs an interaction that makes somebody want it nearby. Recognising the speaker and responding appropriately could help, but the family decides whether that is enough.
The carmaker brings automotive engineering and manufacturing resources to the project, including supply-chain and quality-management experience. Somebody inside that business already knows what happens when a customer calls because something has broken. Those skills could support a household robot, provided the service is organised around the people using it.
For a consumer-services company, setup and repairs are possible areas to investigate. Software updates create another ongoing responsibility. The global pre-orders test interest beyond China; they give us no delivery date for every country.
A factory becomes a training environment
Another automotive supplier is approaching robotics through the information a machine needs about its surroundings. A Chinese lidar company has introduced a handheld scanner connected to a cloud platform that builds three-dimensional digital environments.
Picture somebody walking through a factory carrying a device shaped a little like a wine bottle. The scan gives a developer a digital version of the physical site to use for training and testing. The width of a passage or the position of a work surface can change what a robot can do there.
I would start a discussion about this tool with the task. Can the robot reach an object after somebody moves the container? What happens when the lighting changes? A company familiar with a particular kind of warehouse or production plant can help decide which details belong in the training environment.
The scanner supplier and the engineering partner connecting that environment to a robot's task have different jobs. That gives a buyer specific work to compare when preparing a pilot.
The next experiment needs physical parts
Then there is the wait between changing a drawing and testing the revised machine. JLC supplies electronics and mechanical manufacturing services. At Shanghai's industrial fair, it presented support for robotics development, including circuit boards and mechanical parts.
A change to a robot hand's joint may require a new surrounding structure. A motor driver may need a different board. The engineering team has to bring the parts together before it can learn whether the revision helped.
An open-source humanoid developer in Shanghai provides an example of how that work becomes accessible. Developers can inspect its mechanical designs and electronics, alongside the software used to train and operate the prototype. JLC has described supplying parts during its development.
For an established manufacturer, that opens a practical conversation about a component your engineers can examine. An open design can also give your team a way to test a change on a physical machine.
What the next test still has to answer
I would keep prototype development and production quality separate. A part ordered for an experiment helps an engineer test an assumption. Repeated use in a customer's factory puts different demands on it. An experienced supplier could help a design make that transition.
During a robotics visit, I would ask to see how the team changes the machine. Bring one component your company could help make reliable, and ask which revision the engineers are struggling to manufacture.
[00:00] Cherry, the Chinese car maker, has opened pre-orders for a robot that lives in your home. It is 90 cm tall and it wants to get to know your family. The interesting part is the car company standing behind it. What happens when the people who build cars decide to build your next household companion? Also this week, a Chinese company that helps [music] cars see the road is building a different business. It wants to turn your factory into a digital training ground for robots. Then we look at a less glamorous part of China's robotics industry. Someone has to make the circuit board and the metal parts. Every time an engineer changes the design that can decide how quickly the whole robot improves. And in Japan, researchers have put a humanoid on a skateboard. There's a useful engineering
[00:49] question underneath the backwards baseball cap. Stay until the end for a robot taking on a professional table tennis player. I would be happier watching that match than playing in it. I'm Thomas Daxon and this is Asia Bits Robotics News. I live in Shanghai. I speak Chinese and I take founders, investors, and business leaders to robotics companies here. Each week, [music] we look beyond the demonstration and work out where the business opportunity for you and your company could be. If you want to explore that opportunity for your own company, you can find our robotics visits in the description. Asia base. Let's start with a small robot from the car company. Cherry wants a place in your living room. The robot is called Mo or Mo in Chinese. [music] It comes from
[01:37] Imoga, Cher's robotics business. The company held its dedicated launch in Wuhou on the 21st of October and opened global pre-orders. You may know Cherry through its cars. Here the product is a small humanoid with a face on a screen and a body designed for interaction with children and older people. At 90 cm tall, it is much easier to picture next to a child than one of the adult-sized humanoids we usually see. Immoa demonstrated conversation, stories, and English learning activities. The robot is also designed to give reminders set by members of the household. Imoga says it can recognize family members and use memories of earlier conversations to continue the interaction. Imagine your child asking it the same question every evening. For a family whether the robots
[02:26] remembers the conversation could matter more than how impressive its walking looked at the launch. Living with a machine means meeting it again tomorrow. Now, Cherry is bridging its automotive engineering [music] and manufacturing resources into this project that includes experience with perception and interaction from cars together with supply chain and quality management work. For me, this is the business angle to watch. A car company already has people whose job is to turn an engineering design into a product that can be manufactured repeatedly. It also understands the less exciting questions that follow the sale. something breaks, a customer needs help, and somebody has to answer the phone. Those capabilities could be useful in household robotics. They still have to be organized around a
[03:14] very different customer [music] experience. A robot in a family home has to earn its place after the novelty wears off. Think about a reminder to take your medicine. A phone can already do that. The robot has to make the interaction useful enough that a person wants the extra machine nearby. Remembering who is speaking and responding in a way that feels appropriate could help. That is the product challenge I would focus on. There's also a business around the robot itself. Someone has to explain it to families and help them set it up. Someone has to deal with software updates and repairs. For companies in consumer services, those are possible roles to investigate as these products develop. Global pre-orders give IMOGA a way to test interest beyond China. They do not give us a delivery date for every
[04:02] country. We should keep the launch and the arrival in customers homes [music] separate. What makes this worth following is the combination of a household product and an automotive parent. Cherry is testing whether some of the capabilities behind a car business can support a robot business. for a supplier that opens a more specific conversation than a general prediction about millions of humanoids. Which part of a product can your company help make reliable? Would a robot that remembers your family be useful in your home or would it be something your children stop using [music] after a week? Tell me in the comments what would make you keep it. Very looking forward to the discussion. Wants to scan the place where your robot will work. Getting a robot into a home is one challenge. [music] teaching it about the physical place around it is another is a
[04:50] very famous Chinese [music] company that makes lighter. These are laser sensors used to measure the world around a vehicle. Its new system Cosmo combines a handheld scanning device with a cloud platform that builds three-dimensional digital environments. The former launch was on the 14th of October here in Shanghai where I'm sitting right now. Picture someone walking through a factory with a device shaped a little like a wine bottle. The system records the space so that a developer can work with a digital version of the physical place. That version can become an environment for training and testing a robot. You can see the appeal if you have ever tried to explain a factory task over a video call. The camera shows what is in front of it. A robot also needs information about where things are in relation to each other. [music] The
[05:37] width of a passage and the position of a work surface can change what a machine can do. Kissa is trying to make that information easier to collect and reuse. The company is also working with robot developers including Galbot [music] and Dexal. The product brings a sensor supplier into the work of preparing environments for physical AI. There are several businesses inside that change. One company can make the scanning hardware. Another one can prepare [music] a training environment for a particular customer. The customer may need an engineering [music] partner to connect it to the robot's actual task. Buying the scanner is only one part of that whole workflow. We have already looked at the robot serving [music] ice cream in Shanghai. You remember this time I want to focus on the tools behind the training rather than show you the ice cream. Again, move gives us a
[06:24] clearer picture of how a company from the automotive supply chain wants to serve robot developers. That could be interesting if your business knows a particular type of site very well. Perhaps you work with warehouses or production plans. You know which details [music] change every day and which mistakes cause trouble. That knowledge can help determine [music] what a useful digital training environment has to contain. A beautifully reconstructed factory is a starting [music] point. The next question is what you can test inside. Can the robot reach the object when the container is in a different position? Can it still complete the task when the lightning changes? Those are practical questions an engineering team can work on. My reading of this is that the market around robber training is becoming easier to divide into specific jobs. You can discuss a scanning tool,
[07:11] an environment, and an evaluation task separately. That gives a buyer more useful things to compare than a video of a robot moving [music] its arms. For anyone planning a pilot, start by identifying the information the robot needs about your site. It may change the kind of partner you look for. If these examples help with your own automation plans, subscribe for the next episode. China's next robot design needs a new circuit board. The training software gets a lot of attention. The next story is about what happens when an engineer needs to change the physical machine. JLC, known in Chinese as Jalitang, [music] supplies electronics and mechanical manufacturing services. At the industrial fair in Shanghai, it presented support for robotics development across circuit boards,
[07:58] assembly, and mechanical parts. A Chinese report published on the 21st of October used those services to explain how robot teams work through repeated design challenges. Think about a robot hand. An engineer changes a [music] joint and the surrounding structure may need to change with it. A motor driver may need a different board. The parts have to come together before the team can learn whether the change helped. Every weight between those steps delays the next experiments. This is one reason I would pay attention to the suppliers [music] behind the robots. They influence how often an engineering team gets to test an idea in hardware. One example is Robbo Party, a Shanghai company developing open-source humanoid platforms. We mentioned the company before in a story about robot manufacturing. This time the useful
[08:46] detail is what an engineering team can inspect and modify. Robo Party publishes material covering mechanical design, electronics, and the software used to train and operate its origin prototype. Its public repository lets developers look at the work behind the machine. JLC has also described its support for parts using during that prototype's development. The prototype itself is older than this week's report. [music] The new reason to look at this is the connection between accessible designs and the suppliers that turn a revision into physical parts. Imagine a small engineering team trying to learn about robot's ankle. It can examine the design, work on a change and order the parts for an experiments. The useful result is the next test on a real machine. That is where an assumption in
[09:33] the drawing meets the load on the joint. For an established manufacturer, there are two possible interests here. You might use an open platform to learn and test a component. Or you might already make a component that robot developers need. A specialist in a small precision parts could talk to teams about the revisions they struggle to manufacture. An electronics company could look at the boards and connections that need to survive repeated movement. Those conversations start with something your engineers [music] can examine. I would also distinguish development speed from production quality. A prototype part helps the team test a design. [music] A part used repeatedly in a customer's factory has a different job to do. The opportunity for an experienced supplier may be helping a design make that
[10:20] transition. This is why a useful visit to a robotics company should include its engineering [music] process. Seeing the finished robot tells you what it can demonstrate today. But understanding how the team changes, it tells you where the partner could contribute. Japan's skateboarding robot is learning to move like a person. Now, let's go to Japan, where the demonstration looks much more like a weekend hobby. A humanoid bends it its knees and moves its body on a skateboard. It's wearing a baseball cap backwards. The researchers have clearly considered the dress code. But the project itself, ATR, a Japanese [music] research institute at Kyoto University with support from NATO, Japan's public research and development agency. They showed the work [music] at the CEK exhibition in October. The researcher
[11:08] uses a recorded human movements as a starting point for the robot's learning. They adapt those movements [music] to the robot's body and train in a simulated environment before transferring the result [music] to the real machine. The research program predates this exhibition. The October demonstration is a chance to see the work in a physical setting. Skateboarding makes the control problem easy to say. The board moves underneath the [music] robot. The machine has to coordinate its body to keep moving and maintaining its balance. In the exhibition demonstration, the robot used body movement to perform a pumping motion on the board. It [music] had a safety suspension above it. Keep that in mind when you watch the footage. This is a research demonstration of body control. For a business audience, the useful connection is learning a physical
[11:55] skill from a human example. Think about how many jobs in your [music] company are easiest to explain by showing somebody the movement. A person adjusts a lead or changes posture without describing every muscle involved. Teaching a machine to do that requires a way to connect the example [music] to its own body. The skateboard is a demanding test because balance keeps changing as the body moves. That helps explain why researchers study it. The result gives them something physical to evaluate after training in a computer. We recently looked at Kavasaki learning a lifting movement through simulation. Here the focus is a different skill using human movement as a starting point for dynamic balance. Together these examples give companies more specific questions to ask about how a robot learns a task. Which part comes from a
[12:42] demonstration? Which part is learned in a simulation? What changes when the machine reaches the actual workplace? The answers affect how you prepare a test with a supplier. I would keep an eye on who can turn these research methods into repeatable tools for engineering teams that could make it easier to add a new physical skill without starting the whole development process again. Choose a movement in your own operation that you could show an engineer during a visit. It is a much better starting point than asking whether a humanoid can do everything. These examples show several places where a company can take part in robotics. Cherry is working on the household product. Kuzai is working on the environments used to train machines. And the engineering suppliers help teams turn another design change into another
[13:30] physical test. If you want to find the part that fits your business, join us for a robotics visit in China. Use the link in the description and tell us which process or product you are working in. We can build a visit around the manufacturers and engineers you need to meet. And now for the table tennis. I promise you this is Ace Sony AI's robot playing a professional opponent. The match footage comes from earlier this year. Watch a rally and follow how quickly the record has to adjust. I would ask it for a friendly game. Then I would make sure someone else was keeping score. This was this week's robotics news. See you next week. Remember to like and subscribe.
