Transcript.

Erik: Welcome to the Industrial IoT Spotlight, your number one spot for insight from industrial IoT thought leaders who are transforming businesses today with your host, Erik Walenza.

Welcome back to the Industrial IoT Spotlight podcast. I'm your host, Erik Walenza, CEO of IoT ONE, the consultancy that specializes in supporting digital transformation of operations and businesses. Our guest today is Mo Shen, VP of Global Key Accounts and Alliances, and previously the VP of Engineering at Geek+. Geek+ is a technology company that applies advanced robotics and AI technologies to build flexible automation solutions for warehouses, factories, and supply chain management.

In this talk, we discussed the rapid growth of the autonomous mobile robot market and the domains next stage of development. We also explored the tech stack system architecture and business models behind the solutions brought to market by Geek+.

If you find these conversations valuable, please leave us a comment and a five-star review. And if you'd like to share your company's story or recommend a speaker, please email us at team@IoTone.com. Finally, if you have an IoT research, strategy, or training initiative that you would like to discuss, you can email me directly at erik.walenza@IoTone.com. Thank you. Mo, thank you for joining us today.

Mo: Hi, Erik, it's good to be here. Thanks for inviting.

Erik: I really appreciate you taking time to talk with me about Geek+, because this is one of the most interesting companies that I've come across. I'd say already a fair bit about Geek+, but I'm interested in getting more into the details today. But before we go there, maybe you can just give a little bit of background and how you actually landed at Geek+. I know you were previously at almost 18 years with Oracle. So moving over to a startup, is it bit of a switch?

Mo: Right. My whole career can be divided to two parts. I started with big companies like Oracle and working on database and middleware and applications. I spend almost 18 years with Oracle. Then I moved to another big company, in fact, financial tech, MSCI. After MSCI, I was thinking about what I should do for my life. So it comes across to me that there are a lot of new opportunities, new companies, they’re just growing so fast. And since I’m in channel, and everybody heard about this channel speed, so I really want to join start-up company.

I actually met with few friends and we got talking and we actually started a company in logistics business. I was in charge of the applications one for the WMS software. And unfortunately, that company didn't work out. But towards the end of that experience, I met with a few guys from Ching Hua University, they had this really cool looking company, is a robotics and they're also solving problems of warehouse operators and logistic operators. So we got talking, and they were very interested in my background experience in software. They invited me to join the company, so that's how I joined Geek+.

At the time we’re pretty small was probably around 200 people at the time, the headquarter is in Beijing, and I'm based in Shanghai. Now, that's how I landed in Geek+. And you mentioned that you were very excited about Geek+ and so was I at the time. And I'm still amazed at how we are able to come up with new products and new solutions all the time. I joined in 2018, so it's been almost four years. But every time I look at the company, it's drastically different about a year ago. And we're coming up with new products and new solutions. And I'm still very excited about Geek+ to this day.

Erik: A lot of startups are focused on pure software, and then they can be highly innovative compared to the legacy companies in the industry because they're doing something fundamentally different. But Geek+, you're building hardware and, of course, the software platform that operates the hardware, but you have all of these companies that are building AGVs, and other automation equipment for decades. And it's interesting how quickly Geek+ has become the model, the company that people are sending out their executives to China to study. So I'm really interested actually in diving a little bit more into how you manage to go through these innovation cycles so quickly, something that Geek+ is now six years old. It would be great, Mo, if you can just give us a quick breakdown of what Geek+ does. So what's the core value proposition here?

Mo: Geek+ is robotics and AI company. So we build this robots, we call AMRs, Autonomous Mobile Robots. Then your home automated sweepers, but it's industry size, it's about one meter by one meter. What they do is they work in the warehouse. Basically, in some solutions, they replace some of the movements of materials between different stations, for example. And there's one particular solution that's been used in the warehouse, it's called goods-to-person solution. So basically, in the traditional warehouse, you have people going out to the warehouse floors and pick up stuff from the shelves.

So with goods-to-person solution, basically, the robots will carry the items to you by lifting up the shelves, and move those to the picking station. Or in nowadays, there’re various different form factors for those solutions. There is also a totes-to-person basically. It's not moving the whole shelf or pallet. It's moving one particular tote to the workstations. And they are also solutions to move very stuff in a vertical space, where we have now a new solution called Voice shadows.

So they work on the stack storage, and they can move items between different floors; those either a totes or palate-based, and they will move items to the workstations in need to reduce lag for laborers, people have to move around in the warehouse. But our robots are very generic: they can be used in a lot of different situations. We're still looking at a new verticals and new solutions for the usage of our robots. So that's in a nutshell working capacity and our products and solutions.

Erik: Was this already a category that existed before Geek+ or was this invented by Geek+?

Mo: Yeah, this is pioneered by a company called Kiva and that was acquired by Amazon couple years ago. Amazon bought Kiva and they kept Kiva for its own use. So we saw opportunity in there. And we started with Kiva-like products, but then we expanded on that. I think we now have many more solutions, including the tote-to-person solutions. Kiva started that trend, now I think it's been very popular among the new generation for warehouses.

Erik: And then there's the question of why would it make sense for new product category to come into the market? Now, is it primarily because of the ecommerce use case, or is it more that the technology now exists where previously it was not possible to have this type of autonomous robotic system?

Mo: Yeah, I think it's a little bit both. Why? I think it's particularly a lot of companies in China market they’re interested in this because of the surge of ecommerce. I was in a customer meeting this morning with a global company, and we talked about this Double 11 cell coming up. The mainstream media, they report on those and everybody knows now.

In China, we have these mega sales, and we just don't have enough people to work on the backend. When people buy things, it’s fun, but there's a lot of work on the logistics of backend side, we’re short on people. I read a report about two weeks ago, there was a Wall Street Journal report there's a shortage of labor in the factories nowadays because the migrant workers, they don't want to go to the coastal cities anymore. They want to stay closer to home and they can find jobs actually to home. So there's this shortage of labor, and nowadays, the younger guys don't want to go to the warehouse anymore.

On the one side, there's a surge of ecommerce business. Because of ecommerce business, and the warehouse operation becomes a lot more complex. The traditional supply chain they were doing, walk in, walk out, they operate on either case level or pallet level, now they have to pick out individual items and ship to the consumers directly. That's driving this operation efficiency. Of course, there's this not enough people to walk out at warehouse, so that's on the supply side.

Erik: Yeah, you have this RaaS, Robot as a Service business model. I think it's interesting that this solution allows you to do that, because a more traditional automation system would have basically fixed assets that are installed in a facility, and therefore, you make the investment and then you try to use it, but you basically have that fixed investment.

In here, within this new type of automation system, you could have 10 robots in a given warehouse, or you could scale up to 50 robots depending on demand fluctuations over the year. Are you seeing big adoption of this business model, especially around times like Double 11? Or is it the market still primarily geared towards more traditional purchasing assets, and maybe subscribing to the platform to manage them?

Mo: The current market I will say, from our point of view, we still see a 90/10 split, majority is still being asset purchase, and they want to know on the hardware on the robots and to operate in their own facilities. But you mentioned Double 11 cells. So RaaS is used as supplemental model to help with demand fluctuations. For Double 11, and we have a lot of customers, they come out to us and they want to deploy a fleet of robots to add on to their existing fleet of robots so they can increase on the demand. That's a none way to operate.

The other way to operate that the companies who are new to robotics and they don't want to have a huge capital outlay to pay for the equipment, so they start with RaaS model way. They want to try this out and see how will it operate. It's easier decision for them because we have a large purchases go soon. I think you get a lot of [inaudible 15:03] company, people want to look at your eye. But with RaaS model is a lot easier because you have ongoing operating expense and you move that money and to deploy that to use RaaS as your operating model.

Actually, couple of RaaS warehouses ourselves. So we use our warehouse as a base to provide services to majority ecommerce services to companies. It's probably first year, Double 11 cell was Geek+ now, and we've been doing this as a way also to improve our products. So the five benefit of no operating a RaaS business is that you get all the data, you get all the information about your robots, and you can also deploy whatever solutions you have. Right now, every year we have a new robot and new solutions come out. We tried those out in our RaaS operation, and that helps us to validate the product solutions before we sell this to the market and to the customers.

Erik: And I remember last time that we spoke, you mentioned that more than 50% of the sales right now are abroad right rather than in China. Is there a significant difference in terms of how the business model works abroad versus in China? Is there more acceptance of RaaS in particular markets like Europe, or does this 90/10 rule apply across all markets?

Mo: Either European or US market, we don't have this RaaS option available yet. One of the reason is we don't have our management team, basically, the operators and whoever is going to manage the warehouse and locally in those regions. We are thinking about whether we should expand that to those regions. But we're taking a cautious approach.

To operate a RaaS model, it takes a lot of capitals to do that. Although we get a lot of venture capitalists, still we want to take a cautious view, and when we start a new business in different regions. The reason we started in China is that because it's closer to our home R&D teams and all that: we can get support easily.

In the European and US market. AsiaPac marked outside of China, pretty much still the purchasing model. For those regions, our first priority it's a high labor cost countries. So for those countries, it's easier to justify the otherwise, people really see the benefit.

Erik: I mean, obviously, capital is a constraint. I know other companies are looking at the potential of partnering with financial intermediaries with some insurance company or a PE firm, somebody that basically wanted to play capital. Do you know if that's something that you guys have explored? I'm just curious on how this model is being adopted.

Mo: We haven't explored that model as well. I say pretty much all, but there's some exceptions. People use financial leasing companies as intermediary to provide RaaS supplemental model. But that's a very small part of the international business.

Erik: So your products, you have now a pretty comprehensive portfolio. Is there also, though, kind of an 80/20 rule, where 80% of them are your traditional goods-to-person or what does the adoption trend look like right now in terms of different categories?

Mo: You're absolutely right. I think there's still this 80/20 rule. But they're different forms and shapes of goods-to-person solutions. So right now, our main product is the goods-to-person solution with rack-to-person solution. But now it is a new trend that we're seeing more totes-to-person solutions now.

At least in people's mind, those solutions can provide additional benefit. And some of the business, they have various small SKU items products, so they can fit in a standard totes. That actually helps the company to reuse some of the existing infrastructure they have, also be able to leverage some of the vertical spaces. When we have the rack-to-person, typically, I think there's a limit on the height of the rack, is probably about 2.8 meters, that's the highest one. So in a standard warehouse, they have really high ceiling, so you're wasting all the vertical space.

When the business evaluates the solution, they have to consider all the cost factors. So one part of it is the efficiency of the operation, the other part is really the rental costs. So they want to leverage the vertical space while still keep the efficiency requirements. So totes-to-person right now is getting popular because we have this partner is 8 meter high, our latest, totes-to-person. So you can pretty much reach out to higher levels, and your storage density is a lot higher. So you can probably store 2-3 times more items with those solutions. And so that's additional benefit to our customer.

Erik: So let's say the moving part of the robot would be similar technology, but then on top of that, there would be a platform that can extend up and down and then basically extend out to grab a tote and in draw it in? Is that more or less how the tote-to-person robot looks?

Mo: That's exactly right. So we have this robot with a similar base. But there's a fixed and then also adjustable infrastructure to allow, there’s arm reaching out, can extend and to 8 meter high and to be able to grab totes from the shelves and then move it to the picker station.

Erik: But at least I've had some discussions with companies about the potential value of having basically a robotic arm that can move around because you might need that function in different places at different times. Have you looked into this topic of basically mobile robotic arms?

Mo: Yes. When you talk of robots, I think people talk about industrial robots, which is traditional robots with robotic arms. Those traditional in warehouse, I think they have those prioritized robots, and they can move the cases around from one pallet to another and things like that. Those are one type of robots in the industry use.

And our robots, we call AMR robots are moving with us. You can take an analogy, so it's like your arm, your legs, and there’re different functions. What's interesting is that you can combine the use of the two and make the robot to do more. And we have this custom design where it depends on really what the customer need for the robotic arms. So we can combine the two and make a more complete solution.

Those solutions they’re very fragmented. There's no standard product, no building mass and no have replicate that in difference places. So it's very customized and individualized to a particular usage. There is one particular combination of the two solutions is when you look at the warehouse operation, especially when you look at tote-to-person solution, you will have individual tote moved to the picker station.

And right now there are a few companies who specializes in using their robotic arms to pick out individual items from the tote. So they have a visual picking system to recognize how things are placed and which product it is so they can basically replace the picker at the station, so basically make the whole operation automatic. We do actually have a use case deployed in one of the customer. They’re using tote-to-person solution to find the totes for the order, and they have visual picking system.

So that includes the robotic arms and also the vision picking software. Those vision picking companies, they’re typically software companies, and they have a camera, they capture the image of the totes, and they do analysis and then they give the same instructions to the robots arms to pick out the items from the tote. If you combine the solutions together, you make the whole picking process fully automated without human involvement.

Erik: Certainly, when I look at this, I see a robot, that's a very compellingly visual part of the product. Most of the technology is in the platform, in the software or in specific components around the machine vision element. Can you give us a quick breakdown of the tech stack behind your core robot?

Mo: So our product solutions they’re pretty complex. There's a hallway component. There's also pretty complicated software components. So robot itself, basically, each robot is intelligent device, has a lot of sensor data, has sliders or ultrasonic sensors to detect the obstacles and things like that. In some cases, the robots are fitted with depth camera or laser so they can detect the environment basically. So, there's a perception site.

You have to use sensor data to calculate where you are, basically positioning of your robot in the big warehouse on the big factory. And then based on the position data and the map, of course, some of the robots can scan the whole floor and scans warehouse or factories. There is a map built out for the environment. So based on the position data and your map, the robot should be able to calculate going from A to B, how it navigates around the pass and also the obstacles. So this is one side of this base technologies for the robotics. Basically, you have a perception, you have navigation capabilities.

In a typical usage scenario in the warehouse or factories, we have multiple hundreds of robots running together. So there's a task management side of things. You have to manage the traffic jams and the pass of the robots moving from A to B, although I think there's a direct pass, or maybe it's better to take detour because the direct passes [inaudible 29:56] or used by some other robots. And there’s multi-aging intelligent management system, we'll call it the robot management system. So basically, that takes a lot of information from the robots and sensor data and all that. And it has also a map, basically, of the warehouse and try to manage what's the most efficient ways of moving things around or perform the task.

Then on top of that, we also have application systems. So when we talk about the warehouse operation, and typically, customer has a warehouse management system. So we have to integrate with existing warehouse management systems because the warehouse could also have a manual operation, so you have to combine the automated solution with the manual operation.

So, there is abstraction of this particular area as a robot manager area as a big location to the opposite upper layer, the WMS systems. So you interact basically by sending pick orders to the picking software, and then once the picking operation is completely finished, we usually associated the picking order with certain post number, so the upper layer system will know this is totally picked out, and this is order ABC. So there's multiple layers of softwares.

And right now, we build this very flexible systems, basically, you can integrate with different software systems through different APIs. We work with a lot of different WMS companies like Mehat, and ASAP. And we also work with some custom build WMS softwares. And sometimes in some cases, a partner, they have a very strong software team that want to talk directly to the robot management software. So, we also have an API layer for that kind of integrations.

Also, our customer has different needs and wants, and they want to have a lot of customization in the operations and sometimes they come up with type of usage for robots, so we have to think about how to integrate everything together in some of the cases where they also deploy different types of robots in the same factory. So you have to think about how we know orchestrate different parts of the operations to make sure everything can operate and coexist.

Erik: On the integration side, so I can understand you certainly have to integrate with WMS, maybe you integrate with ERP or other high level systems. Do you ever have to integrate with Edge level equipment? So, would there be some other third party equipment that you would have to integrate with? Or how is that manage?

Mo: There’re some equipment in the warehouse. Sometimes in the warehouse, they have various different types of automation. So they could have a conveyance. They could also have a good coalesce, so we have to interact with those equipment as well. So we do build this layer what we call the WCS layer to interact with different types of equipment in the warehouse or factory.

Erik: In the WCS, that’s an internal layer that you guys have built, or is that part of a WMS?

Mo: Yeah. So WCS stands for Warehouse Control System. So typically, it interfaces with a lot of different automate equipment. And so as the name implies, it's mostly a control software, so controls equipment, talks to the PLCs and tries to make the vertical lifts go from floor one to floor two, this is why instructions it sends to the PLCs. And it also can coordinate and orchestrate various different things to work together.

So, in the case of robotics, sometimes if you need even multiple floors, and say robots traverse from floor one to floor two, so what happens is that the robot will go to waiting point before the vertical lift. And they sense then, WCS will send instructions to the vertical lift basically tells the lift to go to flow one and open up. Then once it's done, it sends the signal back to the robot, the robot can go ahead and move to the lifts. And then the lift over close and move to the second story, then opens up and tells the robot to go ahead and will then go to the destination.

Erik: And on the compute side, on the one hand, you have a lot of real time activity, so this will be done on premise. Is it kind of a hybrid on premise cloud solution? Or is it often just fully deployed on premise? What does the deployment look like?

Mo: We support both type of deployments. For the warehouse operators and from majority of them, there's still apps for the onsite, on premise deployment. But there are also some companies, they want to use a cloud deployment because it integrates well with their existing infrastructure. And also, we're collecting a lot of information data from the robot itself. And we build this data platform, basically, to analyze some of the sensor data we collect from the robots. And we use that as the information data source to predictive maintenance.

Basically, the robots, sometimes even for some of the sensor tags we place in the warehouse, they could have a failure. We want to be able to detect those failures and do predictive maintenance before it breaks down and we can fix those issues so basically, to ensure the warehouse operation can run continuously without interruption.

We do have that components, and that actually integrates very well with some cloud technologies. And we try to use that to supplement our existing warehouse installations. We will also support kind of hybrid. We have a data platform can be running on the cloud. And also, since a lot of robot-control instructions and the software needs to have a real time fast access to the network of the robots, so we tend to have this edge deployments for warehouse management systems that have business applications running on the cloud.

Erik: I have a lot of people ask me about 5G use cases for manufacturing or industrial. And generally, I don't see so many in operations today. So it's a little bit hard to pinpoint where 5G actually will be useful. But on the other hand, you seem to be already basically operating the system quite well with the existing connectivity infrastructure. Is there any advantage that you would see, so with 5G you'd be able to improve performance in some way or is it just more of a cost calculation? How do you guys view 5G internally?

Mo: We do have some early more of a demo cases using 5G in a factory or warehouse settings. I haven't seen a wide adoption of 5G especially private 5G or private 4G solutions in the warehouse or factories. Once the reason maybe, as you mentioned class, I think for 4G deployments, also for 5G deployments, you need to have really have those additional signals in the warehouse or in the factory. So the additional tower needs to build within the existing infrastructure. Right now, 5G is still very expensive.

The benefit brings it's not that obvious. Right now, we are using mostly 2.4G WiFi, or 5G WiFi. Mostly works, only in certain cases as a signal problem, and then we just need to adjust the Aps or whatever the physical infrastructure. We have built this detection process in our robots basically to figure out where you have weak signal, so amplify or something, so basically, make sure those problems go away. That's the only place I've seen of 4G or 5G may have advantage. And it’s less prone to interference from other signal sources. But other than that, the WiFi solution is still more cost effective.

Erik: Well, maybe we can walk through a case. I mean, it'd be interesting to know the before, during, after of a deployment, the one that comes to mind?

Mo: Yeah, we have a huge number of users cases for, goods-to-person solution to analyze one big sports retailer we started out in China, now they have adopted solution and deployed to pretty much all their warehouses in China and plus, in Europe and North America.

So what they have seen is that, after we deploy the solution, they were able to see 2-3 times improvement in terms of the picking efficiency. So if they used to be able to say no pick out maybe 50 orders per hour, now they were seeing maybe 200-300 orders per hour. Because of that, they're able to cut down. They used to outsource their operation to a third party basically just to find temporary labor. They're the warehouse operator to do the basic picking process. Now, they’re able to cut down spendings and the temporary labor's by huge percentage, and they're able to shrink down the workforce.

When we started out, they have over 1,000 people across different warehouses. Now they're able to bring it down to a few hundreds. It's less than 1,000, now I'm pretty sure they keep continuing doing those cost savings. And so this is one example. For sports sweeter, they're doing ecommerce and also there’re store fulfillment operations in the warehouse.

Erik: I guess when you deploy this, you have to restructure the floorplan to some extent because instead of people wandering out, you have a segmented section between where the robots are. Is there a change in terms of the efficiency of the inventory storage? Or what does that typically look like before and after?

Mo: The labor savings is once the main benefit you can get out of deploying robotic solution. There are other advantages: you get a much higher accuracy in terms of the operation. People tend to make mistakes. We have them walk around in the in the warehouse floors and pick out items from just aisles of racks. But when you only have one rack brought to them in their workstation, they're less prone to make picking error so they can improve the accuracy.

Also, I think the training, because they have a lot of temporary labels and if they have to train them know periodically. A lot of accurate them to do and with robotic solution, it's pretty easy. And one of our customers CEO come to the warehouse and they want to learn how this operates. So it's a lot of trainings and it was in typical operation where you basically want to spend a lot of time to train employees.

And they also say because of the save on the space because of the manual warehouse when compared to the automated warehouse, you typically tend to leave a wider aisle because people have to move. And especially for Double 11, you have solids of people in the same warehouse, you have to leave enough room for them to move around. So in the automate case, robots only use very small space, the aisles can be narrower, so it basically get higher storage density. And with the totes-person solution, you can even higher storage density.

Erik: So, Mo, you've given us a good overview so far. Is there anything that we didn't cover that important for people to understand here?

Mo: No, we covered a lot of ground. Thank you for inviting us to join this podcast. And that's an interesting experience for me. Thank you.

Erik: Just then last question, Mo, if folks who are listening, want to learn more about Geek+, what's the best way for them to get in touch with the team?

Mo: I think they can go to our website, www.geekplus.com. There's a lot of conventions they can also contact our sales teams for there.

Erik: Okay, great. Thank you, Mo.

Mo: Alright, thank you, Erik.

Erik: Thanks for tuning into another edition of the industrial IoT spotlight podcast. If you find these conversations valuable, please leave us a comment and a five-star review. And if you'd like to share your company's story or recommend a speaker, please email us at team@IoTone.com. Finally, if you have an IoT research, strategy, or training initiative that you would like to discuss, you can email me directly at erik.walenza@IoTone.com. Thank you.