Interview with Matthias Kohler,
ETH Zurich and Singapore ETH Centre
ROBOTS (Per): Today we are going to speak about automated construction and automation in architecture. So welcome. Could you tell me a bit about how you got into the automated architecture thing. We might talk about architecture and robotics separately later, but how did you get into the combination of the two?
Matthias Kohler: Yes, at the ETH Zurich we started doing research on architecture and digital fabrication, and that involved the question of what kind of research equipment that we would need to do this research. And so we thought instead of buying specialized equipment such as a mill or laser cutter, which very much limits you in the kind of scope of work that you do, we wanted to invest in a more generic machine, a machine which can be customized and adapted to do various different material tasks. And that is how we got into looking into robots, and we finally decided to buy an industrial robot on a linear axis, which allowed us not only to access a big work space, but also to come close to something which is very close to the nature of architecture itself. It is a fact that this machine we could now also assemble parts, we could additively build up structures and if you look at architecture this is very much a feature of architecture, that it is added from smaller parts which then start to define a space and make our living environments.
ROBOTS: I see that. That is very interesting that we assemble from small parts, versus the machining from a big block to a smaller part, which is just totally different…
Matthias Kohler: Correct.
ROBOTS: So we are very well versed in doing automated production for one, but we have no clue on how to do it in this other spectrum where we assemble many small things. What is your research aim now, where are you going?
Matthias Kohler: It is very much about finding out how those new production techniques, such as robotic fabrication, do actually influence the design of architecture itself. So on the one hand that leads to new robotic developments and new material processes. On the other hand that also leads to changes in the culture of how we design with those processes, because each time, in the history of architecture for example, a material system changed, the design of architecture would also change and would take into account the new possibilities, the new performances, for example of a new material system. Now we think, with robotics and the digital computational techniques that are at the hands of today’s designers, there is a very substantial change in the design of architecture.
ROBOTS: Like we see in buildings from which time they come, how they’re built, like a Gothic Cathedral has one shape, a brick building has another shape, a wood structure has another shape and you are looking into what kind of shapes, what kind of possibilities do we have with this new way of doing things?
Matthias Kohler: Yes exactly. We are actually talking of something – the term that we use is a kind of digital materiality – it is the thesis that today matter that architecture is built from, or the materiality of architecture, is not the same as it used to be. It is not just the materials and the way that you assemble them but those materials can get enriched by digital information today, so concrete wall is not just a concrete wall like it used to be; now it can be refined by interwoven features within the building process which are actually features that have come from computational control, or from the robotic processes used, so we think that the very kind of material substance of architecture is about to change, which then leads into new questions of architectural materialization and might even lead to new model systems or different concepts that will be our future.
ROBOTS: What kind of issues do you think that we are going to face? Modular architecture has been around in different shapes or forms, especially in the experimental field for quite some time. Can you tell us a bit about the old way of doing modular construction and automated construction, and how we can move forward, because right now we are building in the old school way still; we are not using automated production? Why did the old school fail, [what are they] and why did they fail?
Matthias Kohler: There are different old schools that I would have to refer to. There is an old school of modular building which did very interesting concepts and some very interesting realizations also in buildings, but usually the idea was that it would be very economical systems that could be easily adapted to different uses, also easily re-configured, etcetera, but what can be seen throughout the history is that usually they have been one-offs as building systems, so they have been used once but then usually not carried on, [so] to some degrees they were also failures. Then a second historic line is the use of automation in construction, there have been for example early buildings in JAPAN where robotic construction was already being tested out, they have been welding steel structures together in high rise constructions, there the idea was basically to replace humans by robots, so increase efficiency and increase safety on the construction site, which worked partially, but the initial costs of development versus the gains in construction were relatively limited. So now what changes today with regards to those two historic developments is that technology today becomes much more software driven, it becomes adaptive, so if we use a robot today, it does not need to repeat the same task over and over again. Like, if you look at car factories, usually the robot does the same movement 24 hours a day. Today it does not need to be the case anymore, so if we look at a kind of architectural robotics, a robot can do an individual task with individual work pieces each time and that is actually through developments of the computational techniques behind the robotic parts, and that is very interesting to architecture, it means modular concepts and also automation concepts in the construction industry can be revisited from a very flexible point of view, till suddenly those technologies can offer large diversity of uses and applications.
ROBOTS: Yeah, because the robot itself is more autonomous, and the digital information already comes from a CAD system, so it can just implement what we have drawn in the computer, [whereas] in the old sense it was just very painstakingly programmed to do one specific thing.
the interface from a CAD environment is very direct
Matthias Kohler: Exactly, yeah. And now basically this interface from a CAD environment is now very direct, so if I draw a line in the CAD program, it can become the explicit information to basically make a robotic movement or to build something.
ROBOTS: Which is borrowing from the CNC routing that you started with, because it reads the module. and then we do this for construction I understand.
Matthias Kohler: Yes, but of course today in architectural design, or more specifically let us say in computational designing in architecture, we go a step further, we not only draw a CAD model as a static model but we also start to literally design the processes that the robots work with, for example we can design behavior of the robots. We recently did a semester where students did a kind of landscape design and the robot was depositing sand, and the students they were not only assigning the robot how to deposit sand, but they were also measuring what formations they built, and had the robot to readjust to their design intentions, so here basically the architect does not design the final forum, but he designs the process, the robotic behavior, which then results in final form which the architect offers, so these are very big shifts for the profession of architecture.
ROBOTS: So that is emergent by the sign the way most of the natural environment we see around us emerges from a very basic simple set of rules gives you any kind of advanced flowers or animals.
We are now at the point in history where we can start to deliberately design those processes. That is quite an exciting phase.
Matthias Kohler: Exactly. And the fact that we are now at the point in history where we can start to deliberately design those processes, so they do not need to be let us say just random bottom-up grown results, but we can actually discuss and design those rules, and see if they perform better or worse you know, simulate that and also build that physically. That is quite an exciting phase.
ROBOTS: Yeah certainly, because it is also so that it is harder to do that in one way but it is also easier to do it in a bigger way, because you do not have to deal with all the small parts in the end, the devil is in the detail we usually say. If we get good at doing these instructions we could emerge very complicated structures and behaviors in quite a simple way.
Matthias Kohler: That is correct. On the other hand we also have to account for the inherent complexity in architecture itself. If you think on how many functions that architecture has to accommodate, it will be quite a tricky task and an interesting field of research also, to see how such rather bottom-up, rule-based design strategies come together with the kind of top down demands that we all have on architecture.
ROBOTS: How do we actually get what we want?
Matthias Kohler: Yes, by designing rules. That is not so straight forward, but that is one of the researches that at least our students and our design courses are doing, for example right now we are here at the Future Cities Laboratory, at the SEC in Singapore, and here the students are designing 40 story high high-rises by designing the rules that built them and they built them those high rises with robotic systems that allow them to build a one-to-fifty model of those high rises, so those are going to be three meter high models which are physically built, and the students have to make sure they stand up, they hold up, and they become interesting in terms of architecture, but they make use of the unique features that robots and the digital control give them in their design.
ROBOTS: This also gives us a scale but still a good test. It gives us experience with answers to certain questions, but it might also might give us new questions, that we need to answer before we go out and do it in full scale. Have you had any of those revelations, where you say that before we go to full scale, we probably have to solve this, whatever it is?
Matthias Kohler: We have a lot of questions popping up almost every day, but it is very exciting and very challenging actually, this feat. I think actually currently the research that we initiated five years ago starts to trigger almost like a field that I would call architectural robotics, so the question how robotic systems and morphologies, how will they have to develop in future so that architecture can make use of them. We are not robotic engineers, we are architects, so at one point this is not our main field of research but I would welcome other people to join this research, with different backgrounds of knowledge. So when we started this research we were very much, we were basically building up a prefabrication cell, which was using a robot, then those kind of imminent questions arose, if we build architecture with robots probably the robot is not going to be larger than the architecture that we build. So probably the robot will still remain smaller which means the robot either stays in the prefabrication, and we have a prefabrication plan very much as we build. Cars nowadays today [?] architectural components, but the other option which we started to look into and which I find relevant also in robotic terms is that we actually have to start thinking about robots which are actually smaller than the artifact that they are working on, so they are actually robotic agents, they are on the building site and are actually building the larger piece on site, they are collaborating to work with material systems so they become like a systemic whole and I think we ventured into this direction quite a bit and that will be certainly one way how those models, the model scale research that we do here in Singapore, will be able to translate. Now there are very interesting paths that we have chosen to look at, there was a recent collaboration with Professor Raffaello D’Andrea, also from ETH Zurich, is to look at building with flying machines. So imagine a kind of aerial construction method for architecture which of course entails a lot of questions, for example about the material systems that you build with, because most probably then you would like to have materials which are lightweight, easy to transport which can fit even if the precision of the position is not so perfect, because in the air, it is quite difficult to get a millimeter accuracy, so that entails a large sum of questions which are highly interesting I think from both let us say an engineering perspective, and that is our main perspective from an architectural perspective, because, most probably, those buildings will start to look different than the buildings we see today out here.
ROBOTS: Many people that listen to the podcast, they come from the robotics field and I think that one interesting question is how can we help you? How can we, what do you find challenging coming from the architectural point of view where do you think we could contribute as roboticists?
It would be very interesting to have robots which are capable of dealing with certain unorderly environments
Matthias Kohler: I think definitely the collaborating agent systems in robotics are interesting, the question how can they orient on a building site. If you imagine a building site, and maybe it is even a bit of a messy building site, maybe there are even humans on the building site. Just thinking of it in terms of flexibility, it would be very interesting to have robots which are capable of dealing with certain unorderly environments, but still can continue on their task, and share with other robotic colleagues what they are about to do, and actually start to assemble a building. What is a very important question to the robotic field is how can they do that at the scale of architecture, because if you look at architecture there is a certain, for example we have room heights which we need to achieve so a room is maybe 3 to 4 meters high, so what could be an adequate robotic system, being able to operate at that very scale, because we find it often that robotic experiments, which are very interesting, are being conducted at a very small scale, and in our view they are not so easy to transpose to architectural scale, and I think that is a tricky question to the robotics people to what degree are those systems and principles they are working on scalable, and how would that be done in a clever, subtle way, without starting to create it like a monster robot, which we are all afraid of, being able of lifting tons of materials.
ROBOTS: So what I’m thinking is that the current industrial robots we see, they are built for a reasonably small scope of operations, maybe a few meters, and they are also built for very high precision…
Matthias Kohler: Yes.
ROBOTS: What architecture needs is a robot with a very large working space but not necessarily with the thousandths of millimeters in precision, so you want kind of the same but different robots, larger working space but you do not require the same repeatability.
Matthias Kohler: Yes, if you look at current building sites you have tolerances of one-to-two centimeters, which are kind of usual. I think it is a possibility for architecture, and for robotics in architecture, to embrace this fact and say well maybe we do not even need to go to the super precise mechanical tolerances, therefore I think that could release the robotics from some pressure in that area but it also means we need material systems that can deal with that. If you have super precise machined parts that you want to fit then they call for zero tolerance right? So that is a lot of development to be done also from the constructive side of architecture.
ROBOTS: One thing we are trying to do there is also to replace the super expensive but very high precision machines with more software, that can read its environment and adapt, rather than be stupid, because that works in the industrial setting but it probably does not work in the architectural setting, because we have a very messy world as you say and we have to have a system that takes in the world and adapts to it and that is also might be better than to go high precision from the beginning because that only works in the industrial environment that is very controlled.
Matthias Kohler: In an ideal setting, in a world which is 300 years from now we could obviously speculate on a fully automated robotically built architecture, etcetera, without any humans on site, etcetera. I am a bit too much of a realist and maybe a bit constrained by that because today I see about 25 professions usually on each building site, cooperating, right, humans, and some machines, and I think it would be very interesting to just start to inject…
ROBOTS: A 26th if you may…
Matthias Kohler: Yeah, a 26th, or a 23rd, or 20th, replace some of the workmanships by robots, slowly showing that they can actually take part in those constructions, without basically having to decide immediately if it is a robotic building site or a human building site, so why not having that run in parallel, and that kind of calls for quite a flexible robotics which can also cooperate with humans.
ROBOTS: Understand humans, and also be safe to work around, and know their own limitations, so the system stops when it does not understand things?
Matthias Kohler: Yes, exactly.
ROBOTS: So I think the roboticists have a lot to contribute to architecture and…
Matthias Kohler: And there is one more topic which I would like to mention, this is more a bit on a technical term, but a big question is also localization, so how do, imagine there is a building site and certain stuff is built [while] other stuff is not built yet, so how do robots now know where they can continue building, what is their reference? Because there is no setup with an ideal point of references which are already available, basically they would need to recalibrate themselves continuously with the information that they already have at hand. That I think it is also quite a challenging task.
ROBOTS: Because then it also comes to optimization the simple thing would be to put one brick on top of the next, and then the next, but that might take an inordinate amount of time. If as you say, each robot can look at the construction site, look at what other humans and other robots are doing, and say if I do this, this is totally, overall the most efficient thing I can do.
Matthias Kohler: That would be, let us say, optimal.
ROBOTS: But if he just could determine that ‘this is what I could do’, I would not be in the way of the humans or other robots and this is a part I could work on, because building sites in comparison to the industry they are very messy, they are unstructured.
Matthias Kohler: Exactly. Also it is one very easy thing to do. You know there are certain things robots are really good at and certain things that humans are really good at, so robots can also make use of the humans that are around, right? So I think that that eases the robotic part a bit, humans can very directly give certain information to robots which would be very hard for robots to find out. But I think in this kind of cooperation, at least in the next ten years, it would be interesting that robotics make an advance, because then they can demonstrate that it is really feasible. If we stick on pure robotic systems without humans, it will be very hard to push it through the reality of architecture. That is a bit my point. That is more like the engineering part, but then of course the interesting question which I would also ask the robotics people maybe to come up with ideas and come to this, okay how they can contribute to the design and the constructive systems of architecture. That might be very interesting, and is also coming up from your fields which might be very inspiring to architecture.
ROBOTS: I guess one thing is natural for roboticists is this active thing. A robot is active; it goes out there; it senses the world and it makes computation and then it takes action. Buildings usually do not do that, they do not sense what you need, and then they do not do computation, and then they do not take action.
Matthias Kohler: Correct.
ROBOTS: You would rather have to go and turn the light on and adjust the temperature and stuff, they do not feel you are hot say ‘hmm, he is probably hot, let us turn the AC up’, so that is maybe a thing that could come from this. Say that 10 or 15 whatever years it takes from now have this, could you tell me what kind of new dream buildings you would build with this system and how they’re different from what we are building today, what is the end product of this? As an architect now, you do not have to worry about the robotics side works just describe the buildings you would build with this.
Matthias Kohler: It is a good question, because this is actually very much of a research question. That is why we have research students, and also us in our office, actually do designs with those technologies. I think those buildings are much more differentiated than today’s, for example, large buildings. Today you often have very homogenous, repetitive buildings, and I think in the future they can become much more diverse and differentiated, while still in their core being quite systematic. They have a logical order to it, yet at the same time they are very diverse; they are a bit like living organisms in the sense of a material formation.
ROBOTS: All trees look basically the same but none of them are identical.
Matthias Kohler: Yeah, for example and I think there are really new expressions, also we have been talking of modular architecture before, I think the question of modularity might come back in architecture again, not in the sense of strict repetition of a modular system but more in actually quite open and flexible modular concepts which create really very different environments, for example the recent exhibition that we did together with Raffaelo D’Andrea was flying quad copters, then we invented some kind of super module, the super module was 13 meters by 18 meters in planned dimension and 10 meters height. It was a huge module, which could accommodate for example a cinema, but it could also accommodate 4 to 6 flats, it could accommodate all kind of different functionalities, and it has been assembled in a way to a mega structure, actually 600 meter high vertical village for 30,000 inhabitants in a very [varied?] way and formation, which makes it really a new proposal on the architectural side yet the concept towards that each of those modules would be fabricated in a kind of field factory on site and each one would internally be completely different but from the outside it would look somewhat the same so I think there are new scales and concepts of modularity that will now be probed with those new technologies. That is just one example.
ROBOTS: Should we talk a bit about like the certification process, because I guess architects are just so used to always have that in the back of their minds, this is the building code for where I am working, I adapt my work to it. I do not think that roboticists are very good at that.
Matthias Kohler: It is true, that is a fact of architecture that you adapt to your local building codes, but there are differences in building codes, there are building codes that are more on the usage of the terra, etcetera, so they are more or less independent of the construction technologies that you use, but then there are certain building codes like earthquakes, etcetera, which really….
Matthias Kohler: Yes, like fire, which we relate very much to the construction.
ROBOTS: Robots have an advantage there, over humans, it is that when you actually taught them to do something, they will actually do exactly that, the quality control is one thing.
Matthias Kohler: There is an advantage and there is a problem. The advantage is exactly what you mentioned that, for example, safety on the job site can actually be improved with robots because you have a protocol operation, you can have less humans on the job site or you can ensure the quality by doing quality assessment while the robots build. You have another kind of problem because classically building materials are certified, that you use in a building, but they are certified for a standard use case, and I think what we are talking about in the future that architecture is going to take, materials are going to be much more adapted and customized, or maybe modularized, that is going to be seen, and there will probably be the need for new kind of certification standards, for non standard systems, which are really heavily, can be well pre-simulated, but which exists only once.
ROBOTS: So what we must do is we must adapt the certification process, so that it could work with the models we build, and actually certify not only the material but also the use of them in this particular construction.
Matthias Kohler: Exactly.
ROBOTS: So that we as architects and as builders and engineers and robotisists know that we are building according to the certification process, which is actually a set of best practices; we know that this is a good way of building buildings that are resistant to fire, or that are resistant to earthquakes or whatever.
Matthias Kohler: Yes, it is basically just a shift from another rather static definition to a more dynamic definition and I think this is a general tendency that we have today, in the digital age, that we start to have more and more dynamic processes, those are also the processes that need to be kept new standards and certifications.
ROBOTS: Yeah, because this is something I think that roboticists are not very used to working with because we do not work out in this. Our industrial robots that we use to design they worked in a very closed cell, if somebody just came close it just shut down and that was it. These robots have to work with people, and the result that they produce has to work for people.
Matthias Kohler: You know we had a very hard time with the robotic machineries that we bought, to explain to people that each time we build something we build it differently, and we would also like to have people sometimes intervene in the building process, etcetera. This is really something the robotic industry at least is really not used to. I think in research we are a bit further, but I think those concepts also need to become a reality, and I am also very positive that those concepts will become a reality and this is going to be a very exciting future, I think, for architecture and for robotics.
ROBOTS: So, as a roboticist, it is going to be interesting to see the work sites, see all the robots work, and of course somebody that has a house, and I enjoy architecture, I am not an architecture buff or anything, but it is going to be really interesting to see what kind of signature buildings, what Cathedrals of Chartres are we going to build with this new technology, and I hope that they we as roboticists are going to be able to help you architects create buildings that really enlarges the scope of humanity, because that is what architects do, they provide space for us to do whatever we need to do.
Matthias Kohler: We very much welcome every help in the field, and I think I am looking forward to really great future research prospects, and I am really excited to find out the architectures that will come out from that.
ROBOTS: Thank you, thank you very much. I am sure we will be back soon.
Matthias Kohler: Okay great, looking forward to that.