00:00:01Dr. Eneko Osaba: When we talk with clients, with our partners, with our colleagues, they heard a lot about quantum computing in YouTube, in blogs, in the newspapers, but sometimes they don't know what we can make with quantum computing. So when we solve these kind of real problems, we just say, "Okay, talk about your problem. Don't hide nothing. Talk me about all the problems, all the constraints, all the difficulties that you have, and then let's see how we can approach this problem."

00:00:36Murray Thom: Hello, and welcome back to Quantum Matters, where quantum computing gets real from D-Wave. I'm your host, Murray Thom, as we move past the hype and the theoretical to explore practical real world applications of quantum computing today, and where the biggest opportunities lie in the future. Let's open the box and see what's possible.

Today, we're going to talk about applications that bring quantum computing together with real industrial robotics. When we think of practical optimizations, our minds might go to routing, scheduling, and resource allocation. But on this episode, we're going to look at how this enters into robotic path planning for quality inspection. This discussion will not be about abstract theory. It's about solving real problems with real robots, where speed and practical performance matter more than theoretical perfection, where quantum computing becomes part of a broader engineering workflow. Not magic, but a powerful optimization tool to make processes faster.

My guest today is Dr. Eneko Osaba, Principal Researcher at Tecnalia. He has completed multiple projects using D-Wave's hybrid solvers, including a quantum optimized robotic quality inspection system with impressive real-world results. Eneko, welcome to Quantum Matters.

00:01:50Dr. Eneko Osaba: Hi, Murray. Thank you for inviting me.

00:01:53Murray Thom: So, can you tell us a little bit about Tecnalia and what you do with them?

00:01:57Dr. Eneko Osaba: Tecnalia is the biggest private research and applied center in Spain. It's also one of the biggest. In Europe, we are more than 1,500 people. Of course, not all these people is working on quantum computing. We work with artificial intelligence, computer vision, cryptography, and quantum computing. So we have a small group, more or less 25, 30 people working on quantum computing and quantum sensing. And I like to say that our strongest part is solving real world problems using quantum optimization, (inaudible).

00:02:37Murray Thom: Excellent. I mean, I think that's what everyone is really excited to hear. Quantum computing as a tool, and then what are the kinds of places where it's being used. So can you tell us a little bit about how you first got started working with quantum computing?

00:02:49Dr. Eneko Osaba: It is a very good question, and it is quite funny because I'm not physicist. I studied computer science, and I make my PhD in classical artificial intelligence. It was more or less in 2020 when one colleague of mine, (inaudible), told me, "Okay, Eneko, why don't we make something with quantum computing?" Okay, so at this moment, I didn't know nothing about quantum computing. So my first reaction was to go to Google and put quantum computing and traveling salesman problem. Why? Because this is my favorite problem. So the first link was about D-Wave, D-Wave solving this problem. So little by little, and learning by myself, I started working with D-Wave and working with this kind of problem.

So this is how I started. It was not as magic as, okay, I studied a physicist and then I make a master in quantum computing. No, this is not my case.

00:03:54Murray Thom: Yeah. Well, I mean, I think that not a large proportion of society start with a quantum mechanics background. So I think that definitely a lot of folks are going to identify with that experience. I want to make sure we begin with how you've successfully implemented a project on quantum optimized robotic quality inspection with some impressive results. Can you tell us about that application and the results that you achieved?

00:04:15Dr. Eneko Osaba: It started with a public funded project here in the Basque Country in Spain. In Tecnalia, we have a lot of departments and we have some kind of facilities and also robots. So this is a problem that they have in one of our departments, in one of our other areas, and it was about the quality inspection in industry 4.0. The thing is that today it is a very demanding field in which all the manufactured parts must be inspected because they have a zero defect requirement. So this department, they have a robot that inspect the piece that is manufactured in order to check if everything is well.

So the thing is that we have one piece that we want to inspect, that piece from the computer aided design. We obtain the map, the map of the piece. And in this map, we set some kind of inspection points. More or less in our research are 100, 200 inspection points, and the robot must visit all must inspect all these points. This is a trajectorial problem because the objective is to minimize the energy consumed by the robot. So thanks to the hybrid solvers of D-Wave, we inspect all the points minimizing the route, so minimizing the energy consumed by the robot.

And something that I like to say when I talk about this use case is that everything is real. The robot is real, is in our facilities, the pieces are also real, are manufactured in Tecnalia, and of course the algorithm is real. It (inaudible) the real quantum computer of D-Wave.

00:06:06Murray Thom: When you're visiting these paths, could they be optical inspections? Could they be in contact with the objects? I mean, because I saw you bring in an example of one of these objects and they're quite complicated three-dimensional shapes. So how is the robot interacting? What's the inspection part of that task?

00:06:22Dr. Eneko Osaba: Yes, the robot has one camera, it has a sensors, and it just put very near from the piece and take a photo. But this is a very good question because the pieces are three-dimensional, so there are some paths that cannot be made. If you are thinking about an aircraft, it has no sense that the robot goes from the tail to the face of the aircraft. So this is why the graph is not complete. There are just some paths that can be completed and other that cannot be completed.

This is very good news in this use case. Why? Because if the graph is complete, the problem is more complex so the problems that we can solve would be smaller. But in this case, we are solving problems with, as I say, more or less 200 of inspection points.

00:07:15Murray Thom: And I think that quantum computing is such a new field. I mean, a lot of folks are just getting familiar with what the objects themselves are, but I think a lot of folks are also curious about how to assess if something is a good fit as an application. I mean, what helped you to recognize that this would be a good fit for quantum optimization?

00:07:34Dr. Eneko Osaba: One of the first things that we make when one client or one other department approach to us is to make an informal discussion, is just, okay, describe your use case, describe your problem. So in this case, for example, some things that we check is that, okay, the variables of the problem are discrete. Or for example, something that is very, very important, the objective faction of the problem is heavy or it is just a mathematical function. In this case, the objective function is just a mathematical function.

Another characteristic is that, okay, the problem (inaudible). We have a complex problem. This is also the case as much inspection points we have to inspect, so the solution and the search space of the problem is bigger. So we are testing all the checklists and if some or all the characteristics are met, this is a very good problem. In this case, this problem is a very, very good candidate. And the results supports our decision of solving it with quantum computing.

00:08:41Murray Thom: So when you were saying there are variables in the problem that are discrete, it's kind of like if you're going to choose a path and you're moving between two points, it's sort of like, am I going to go to point A next or am I going to go to path B next? Those are kind of decisions that either happen or don't happen. That's what you mean by discrete. Is that right?

00:08:57Dr. Eneko Osaba: Yeah. When I talk about discrete, this is a very simple example, okay? But if you have a route of just 10 points, the variables are discrete. Why? Because when you visit one of the nodes, you just have 10 different options to go. So this is a discrete solution space. So in the first visit, you can go to one of the 10 nodes, and in the second part, one of the nine remaining nodes. So this is not a complex continuous mathematical function, it's a discrete solution space.

00:09:34Murray Thom: The next point that you mentioned that I just wanted to expand on as well, as you were talking about a mathematical formulation. So for instance, a mathematical formulation might be like, if I choose point A next, there's a certain cost associated with that. If I choose point B, there's a certain cost. And then you look at all the routes to sort of figure out, okay, well, how do those choices affect the routes that I can take? But you also mentioned what's great about that is that it's not a heavy function. So expand a little bit on what would be a good example for people to think about for a heavy function?

00:10:01Dr. Eneko Osaba: We can think about a smart city. This is just an example. You want to put traffic lights. If you have one solution, one possible thing to test if your solution is good, it's just to make a simulation, a very heavy simulation, maybe 10 minutes seeing if all the cars are going well, no incidents. So this is a heavy function. But in the industry, most of the problems, they have a mathematical function, a very easy, very fast to calculate mathematical function. And this is the case, because when you have a route, it's easy to calculate how much energy the robot needs to go through the route. So this is when we can distinguish between a light objective function and a heavy objective function.

00:10:53Murray Thom: One of the things I'm curious about is what were the results that you saw? What kind of outcomes did you get from this project?

00:10:59Dr. Eneko Osaba: When we thought about this problem, when we design experimentation, what we want to make is to make a good impact. And if we want to make it, we need to compare the results with a classical baseline. So we choose two of them, the Gurobi Exact Solver, which is one of the baseline algorithms to compare. Another one is Google OR-Tools. It's an algorithm then designed by Google, so it is a good one.

So we made a comparison with these two algorithms and the results were quite impressive. We use different hybrid solvers of D-Wave, but I'm going to focus in the Stride one, which is the best one for this kind of trajectorial problems. It is better in both quality and runtimes in comparison to Google OR-Tools. And if we compare with Gurobi, okay, Gurobi in terms of quality is better because it is an exact solver, but Gurobi needs a lot of time, lot of run time. For one instance, it needs 60 minutes to calculate the solution, but the Stride solver just need five, six, seven seconds. And this is something that in the industry is very, very valuated.

The problem is that when we think in a real problem, we usually think that, okay, my client wants the optimal solution, the best solution. But no, no, this is not always the case. We have many projects, many clients that, they just need a very good solution, but very fast. So this is a good example because if we need to check, I don't know, 24, 30, 50 manufactured piece by day, we cannot wait 60 minutes for have one solution. So in this case, we need very fast solvers. And this is the case in which the quantum hybrid solvers can have something to do.

00:13:01Murray Thom: Let's explore that a little bit. So you're able to use a really powerful, well-researched and well-studied solver like Gurobi to establish a baseline and get a bit of a sense of the quality that you can achieve. So then in that scenario, you would use, let's say, Google OR-Tools or the D-Wave's Stride in order to figure out, well, how close to that optimality can you get? And how do you measure close to optimality there relative to the Gurobi solution?

00:13:26Dr. Eneko Osaba: It is just the energy consumed in the route. So we assume that Gurobi reaches the optimal solution of the problem and we calculate how much energy needs the robot to build or to make real this trajectory. So we calculate the same for the routes that we obtain, both from Google OR-Tools and from the Stride hybrid solver and the approximation ratios are quite impressive. I mean, it is just five, six seconds and the energy is just 15% near to the optimal one. And in these industrial use cases, this is enough. This is a very good solution.

00:14:06Murray Thom: Right. So you're getting within 15% of optimal, but instead of running for hours or multiple hours, you're getting it down to five seconds. So you're getting kind of that speed of a fast tool, but much closer to the quality of these really powerhouse longtime running tools.

00:14:22Dr. Eneko Osaba: Yes. This is indeed the worst part of the Gurobi or other heavy algorithms that okay, they can reach good results, but they need lot of time. Lot of time. Maybe in some industries, they don't have this problem, but the problem of the runtime is something that is more important than the problem of the quality of the solution. And in this case, we are talking about just five seconds in very big instances and 80, 90% of the solution quality in comparison to the optimal one. So it is a great, great results. They are great results.

00:14:59Murray Thom: Oh, congratulations. That's fantastic. I love the fact that because it's not an abstract problem, you have a robotics department, you've got an opportunity to speak to the folks in that department to really understand the nature of the problem. I mean, how much does that change the approach of the optimization formulation when you're talking with the folks who are doing the work themselves?

00:15:21Dr. Eneko Osaba: I can talk a lot about this, because when we talk with clients, with our partners, with our colleagues, they heard a lot about quantum computing in YouTube, in blogs, in the newspapers, but most of the times they don't know what we can make with quantum computing. So when we solve these kind of real problems, because in the literature, there's a lot of academic problems that are solved. And this is very good. We also solve academic problems, but we are talking here about real problems. And when you talk with one colleague, with one client about the use cases that you are solving, they are very easy to say, "Ah, okay, you are solving that. You can do this with quantum computing. Let's talk about my problem. Let's talk about my restrictions." When we talk with our colleagues, with our clients, we just say, "Okay, talk about your problem. Talk me about your problem. Don't hide nothing. Don't hide nothing. Talk me about all the problems, all the constraints, all the difficulties that you have, and then let's see how we can approach this problem."

00:16:28Murray Thom: Well, I love this, because it reminds me of something I'd written about software and hardware in a book written by Nancy Leveson about engineering a safer world. And she kind of really identified something interesting about software development, which is that software development allows you to design ideas without having to worry about their implementation. And that is, I think, kind of really helpful to think about. And what you're sort of saying is that in these cases, when you're solving real world problems, you don't want to move in that direction, because once you lose the details of the implementation, you're actually losing a big aspect of what makes them complex. And businesses, that can sometimes be the root of the gap between academic solutions and the solutions that businesses need. So if we want to be able to take these technologies and help to realize that value and that speed in an industrial context, those features of like, am I cutting wood with a saw or am I machining aluminum with a heavy metal lathe, those differences really matter in terms of those constraints.

Now, I have a question for you, which is that I think that what's sometimes on people's minds is the idea that quantum computing must be really complex, and then implementing solutions might take a long, long time. So can you tell us a little bit of what was the effort that went into the implementation of your solution here?

00:17:51Dr. Eneko Osaba: This solution, I think that it was more or less four months, talking about the data, talking about (inaudible), the implementations, talking about the benchmarking, because it is hard to obtain the problems because you need to extract the computer aided design of the manufactured piece. So four months, making all the preparation, the implementation, also the checking, the evaluation, also implementing and running the classical counterparts of the CQN at the Stride hybrid solver. But more or less, more or less, I think that it was four months.

00:18:28Murray Thom: I think it's fantastic because when folks are working in industry and they're looking to implement new software solutions in their business, four months is a short period of time. It's sort of that discussion you're talking about like, okay, well, let's make sure that we capture the data that we need to be able to incorporate the solution. Let's talk with the experts, make sure we've captured all the constraints. That is like a software development kind of workflow. And if you can do that and incorporate quantum computing and then start to create these advanced solutions in timescales of five seconds, what's phenomenal is that people are starting to recognize that quantum computing maybe unexpectedly is removing complexity from their business. It's not adding complexity if you're able to build these applications on a familiar timescale for them.

00:19:11Dr. Eneko Osaba: Yes. Something really important is that people think that quantum computing is very, very, very complex. Yes. And indeed it is a very complex field, but not as complex as the people think. And I think that one of our responsibilities in Tecnalia, it is not just to solve use cases. It is to talk with our clients about what quantum computing can do and what quantum computing cannot do to make quantum computer real to our clients, to our colleagues. Something that they really like and something that they usually say is that in Tecnalia, we take quantum computing from the laboratory to the industry, from the laboratory to the real use case. And in this case, is when you can talk with clients, you can talk with your colleagues, you can build the path together. So I think that this is something very, very, very, very important.

00:20:04Murray Thom: Well, and I think that's really fascinating from the perspective that we are building quantum computing, quantum computing is a tool, but the reason we're building that tool is to help people to be able to get a benefit from that. And at Tecnalia, if you're talking with your enterprise clients and customers and helping them to say, "What challenges do you have in innovation? What kind of problems do you have? We have understood this technology and we can help you place it there," that's playing a key role in the economy, because that's going to allow to help those businesses to innovate more quickly. It helps them to reduce the perception of the risk and the time that it's going to take. So I think it's a really key piece in the overall position, and it must be really exciting. I mean, the culture of Tecnalia must be thriving from that perspective of working on that innovation.

00:20:51Dr. Eneko Osaba: Yes. One of our lemmas is we search problems. And this is something also related with the things that I say before. We want to solve real problems. Okay. When I talk about academical or academic problems such as the TSP of the job shop scheduling problem, okay, we solve the jobs shop scheduling problem, but not just the academic version of the job shop scheduling problem. We solve this problem with flexibility, with machine availability, and based on client orders for example. We solve the TSP, but not just the TSP. We solve the TSP with the time windows of the clients, with pickup and delivery. Okay, we solve their impacting problem, yes, but we solve it with incompatibilities, within different packages, also fragilities. And this is how little by little we can give answer to the needs of our colleagues and of our clients.

So in a few words, the philosophy of Tecnalia is just, okay, we search problems. We want problems.

00:21:56Murray Thom: I love it. I love it. That's great. Now, a lot of folks who are looking at quantum computing might be early in their journey. Maybe could you share what's one piece of advice that you would've given yourself when you first started out in the quantum computing domain?

00:22:11Dr. Eneko Osaba: I think that I have a good advice based in my background. Okay, I am computer scientist. My PhD is in classical artificial intelligence. I love to think that I am making good things in quantum computing. Okay, quantum computing is a complex field, but not as much complex. So please don't be scared about quantum computing. If you want to start with quantum computing, start and read and download a code from GitHub and start implementing things, because it could be easier than you think. Look at me. As I say, I am not physicist. I am computer scientist and I am making quite beautiful things with quantum computing. So this is my advice, don't be a scared.

00:23:01Murray Thom: Yeah. Get started, approach it. Don't wait. That's a great point. So Eneko, it's been great having you on the podcast and having a chance to learn about your amazing work and these practical applications you've done with D-Wave technology. So thanks so much for being a part of Quantum Matters.

00:23:16Dr. Eneko Osaba: Thank you. Thank you so much, Murray.

00:23:21Murray Thom: That was really fun talking with Eneko. And I'm excited to tell my friends I've recorded a podcast episode that brings robotics and quantum computers together. I mean, this scenario of getting within 15% of optimal, but instead of taking hours to get there, doing it in five seconds, I mean, that really makes an application possible that wasn't before when the scenario is that you have lots of different parts to work with and dynamic planning is an important part.

It's also great to hear that quantum computing is really delivering on removing complexity from business rather than introducing complexity when the application development process is familiar about getting access to data and working with experts to make sure that you've captured the real problem and delivering in a familiar timescale. And I love the way Eneko finished. Quantum computing is probably a lot less complicated than you think when you get started, so you should start right away and it's a lot of fun to work with.

That's it for this episode of Quantum Matters. So thanks to you, our viewers and listeners for joining me. Please follow so you don't miss an episode. And to learn more about how D-Wave works with organizations to get started and succeed with Quantum, visit dwavequantum.com. Until next time, I'm Murray Thom. Stay curious about your quantum reality.