In an interview with CIOTechOutlook, Dr. Prince Joseph, Group CIO, SFO Technologies explained how startups can leverage AI, machine learning, and cobots to revolutionize small-batch electronics manufacturing. He underscored modular design, open-source technology, and interoperability. Additionally, he points out the necessity of digital twins, collaborative development, and adaptive integration to reap long-term value.

What challenges do startups face in integrating AI and machine learning into cobots for small batch electronics manufacturing?

The whole concept of small batch production, especially when looking at SFO's position, one of the things that differentiate us from other manufacturing entities is the type of work that we take up in manufacturing, where the emphasis is on low volume but high mix. This means a wide mix of components, technology, and engineering involvement, while the quantities made may not go into the millions, but only into the hundreds or thousands.

The main difficulty that startups face is that, for a robot to work effectively, it has to be programmed to do specific tasks. Although there are some tasks that, irrespective of the product, can be deployed with small changes, the question is: are we able to figure out how to find the right use case for deploying such a solution? And can this be used for all the variety of products that we have to manufacture? What are the prerequisites and inputs that are needed to make this successful? We are currently concentrating on the data that needs to be available. In certain instances, the quality of data could be good; yet in most environments, the quality, distribution, and acquisition of data—and integrating it to develop a robust model specific to small batch production—are huge challenges.

The introduction of digital twins can help with the data flow where the production environment is simulated. One area of focus is cobot, which is not doing very specific repetitive tasks because it's a robot, instead, it's about intelligent systems that can respond to a variety of parameters and perform functions accordingly. It is important to have some critical feedback loop that keeps the learning process continuously. Testing validation and monitoring are constantly going to be important.

In general, Operational Efficiency (OE) is increasing. While these technologies can lead to cost savings, it is important to extensively research the particular context beforehand. The main challenge is if it is small-scale production, then the data is available for you to feed, to train, to create the model that will constantly relearn and having a mechanism for the feedback loop.

How do startups differentiate their cobots from those offered by established industrial robotics companies?

The startup environment today offers a tremendous amount of opportunity in what they can accomplish. For big players, the largest obstacle to adoption is that it might be perceived as risky because of the cost aspect. Also, it is tailored for such an environment that it might not fit with the operations you are attempting to perform or the routing layout in your plant. Hence, there is a lot of risk involved in taking up proven industrial robotic solutions. Startups can enter the space and provide pilot implementations in a context-specific manner with much more affordability.

The other advantages that startups can provide are flexibility in design, plant layout, integrations, and post-sales support. Tailoring a solution for a specific establishment—whether for a small-scale industry, an operation, or manufacturing—is best done by startups that can work in a collaborative way, similar to cobot, jointly to co-develop the solution. In developing a cobot or any industrial automation, it is always developed for particular contexts. That environment must be mature and clearly defined for the solution to work and yield results. Either the whole operation must be adjusted to accommodate the solution, or the solution must blend in perfectly with the operation—and figuring out which method is more effective is important.

What funding challenges do affordable co-mode startups face and how do they navigate them?

There is high amount of R&D and engineering involved. There is a lot that goes into this in terms of design, and it takes very much grassroots personnel who are seriously working on the project. That value is not appreciated by typical funding providers.

Deep tech is one such sector with huge potential, but there is also a huge risk involved. Not all that is created going to work. There needs to be collaboration with venture capitalists or tech accelerator investors who at least have some idea about this space and can help in making the right connections and networks. It also gives the liberty to create solutions that can be taken to market.

I think the way the startups are working now, either they are way too small or they are not able to find the right funding with the right idea and not talking to the right people. This is something that needs to be developed. The conversation in this space needs to improve. Only then can the whole sector feel a positive impact. Startups need to work with the right sort of funding partners, and funding providers, and leverage the right sorts of government schemes, which will enable this sort of a conversation.

What are the key technological challenges in designing cost effective and precise cohorts for small scale electronics?

For small-scale electronics industries, there are some very unique requirements. For example, one of the cases that we were working on, required some parts to be machined, can we even consider 10 microns or 20 microns? So we're not talking millimeters, centimeters, we're talking microns. So how do you get that level of precision to be done? If you're looking at a large-scale automation solution, it may be possible, but bringing that in for this sort of machining, which may be only needed for a small quantity, is just not going to be the right solution. So either the task is contracted out to a party who possesses that ability, or it is constructed so that what is wanted may be produced using a design particular to the environment.

Approach the Design in a modular way, consider context, and use selected open-source technologies, since this will reduce costs, support interoperability, and reduce risk to a lesser extent. Opt for precision and efficiency. When applied to product development and mass production manufacturing operations, this strategy will provide the optimal long-term value.

How do startups ensure cohorts integrate with legacy manufacturing systems without major infrastructure changes?

This provides an opportunity for startups, as established large players, by default they tend to become proprietary in nature. The manner in which their systems are designed and built necessarily that they have to work within a given context. Whereas the startups have an opportunity to build in that interoperability part in the design of the cohort itself. So the startups know there are a variety of IoT bridges that can be thought about.

There are some protocols—Modbus, MQTT, OPC, and others—that can be taken as the foundation, and they have built-in support right from the start. The system is also compatible with a standard set of MES, CRP applications, CSPs, and Microsoft platforms. If there is a machine that cannot be integrated directly, we have a component that can be plugged in which is an IoT device. From there, we can create a means of communicating with the machine. If it is a machine that already has communication capabilities, then we know what protocols support it.

We can consider integrating it immediately. In most legacy domains—like CNC machines and certain legacy PCB assembly lines—plug-ins from the outside might be necessary. But the newer Industry 4.0-graded machines are built for connectivity, so integration is simpler. Although risks are present, it is crucial to perform an analysis of the environment where the component will be used and then look at the integration needs. This provides insight into good uptimes, and the investment pays off right away.