建築情報学会_定期学術会議_ラウンドテーブルセッションまとめAIS_Annual Academic Conference_Summarization of the Round Table Session




The application of traditional knowledge and techniques in digital construction. -Towards Crafting the Digital




  • Mr.Miyaguchi,M (Robotics Restacher,Takenaka Corp.

  • Mr.Kawakami,T(Structural Engeneer, Takenaka Corp.)

  • Mr. Takasaki,S (Master capenter, Hakuho-Shaji)

  • Dr.Xie,h (AI-UX researcher, JAIST)

  • Dr.Hotta (C-Architect, Keio University)

❷登壇者(敬称略 )

  • 宮口幹太氏(ロボットG, 竹中工務店

  • 川上拓馬氏(構造家, 竹中工務店)

  • 高崎将太郎氏(宮大工棟梁, 白鳳社寺)

  • 謝浩然氏(AI-UX研究者,北陸先端科学技術大学院大学)

  • 堀田憲祐氏(Cアーキテクト,慶應大学)


In the 2020s, the technologies such as learning machines by themselves and mechanisms that mine data by themselves became commonplace. Using those techs may be possible to extract valuable essences from "things" or "behaviours". Hopefully, we will be able to efficiently learn the human wisdom that humans have accumulated so far. Here, I would like to discuss what is possible by digitising traditional architectural techniques in general and their applicability based on case studies.



Case Study

In the coresearch between Keio Ikeda-lab and Takenaka Corporation, the Researcher studied what and where the professional-master carpenter is looking at/ observing in the process of wood marking. This study is made with language protocol analysis. In addition, the observation elements extracted in a rational and element-reducing manner were reconstructed and mounted on the robot arm. Despite trial and error, we have partially succeeded in improving various physical properties and the ability to grasp the object's state to be processed.


➀慶應義塾大学池田研×竹中工務店の一連の共同研究(Sukegawa.C et.al 2022);木取りの過程において、「宮大工さんの目」=熟練技術者は作業の際に何を、どのようにして観察しているのかについて、言語プロトコル分析を通じて詳細に書き出した。またその推論的・要素還元的に抜き出した観察要素を、再構築してロボットアームに実装した。試行錯誤はありつつ、加工対象物の様々な物性や状態把握能力を向上することに、部分的に成功した。

Palace carpenter research (Keio University x Takenaka Corporation, image source taken by the author)宮大工さんリサーチ(慶應大学×竹中工務店、画像ソース筆者撮影)

2) Running coaching science (Pro. Iso / Waseda University); this reseach gave an example of using digital technology in sports science. Using motion capture technology, we scientifically and objectively proved the knack of how to sprint a player with a record (especially regarding posture) and used it for coaching sprinters.


Running coaching science (Pro. Iso / Waseda University)
Quote https://www.youtube.com/watch?time_continue=2&v=hrZlbdBWiSc&feature=emb_logo)

走りのコーチング科学/ (磯教授/早稲田大学)

引用 https://www.youtube.com/watch?time_continue=2&v=hrZlbdBWiSc&feature=emb_logo)

3) Dynaformix (Shinichi Furuya / Sony CSL); Wear the device attached to the finger of a professional pianist, the researcher let them play the piano. Thus, 1) record the skilful fingering and 2) output it directly to an unskilled person by force feedback, are executed. It showed the possibility of saving essential practice time. By doing so, it is possible to learn such things as the usual way of taking the time of the expert, fingering, etc., with physical re-experience.


QuoteDynaformix Furuya Shinnichi(Sony CSL); https://www.youtube.com/watch?v=l2jkkkVXjo0)

画像引用;Dynaformix古屋晋一(Sony CSL); https://www.youtube.com/watch?v=l2jkkkVXjo0)

❺ Hypothetical / discussion

1) E.g.How to learn; A craftsperson who implicitly knew the behaviour of his master, a new generation who discovered the method by teaching methods such as language and pictures, and a search generation who searches videos and source code to find solutions efficiently. Will the age of true digital natives change further?



Is it possible to "mining tips" by digitising the behaviour of skilled technicians?

→ Isn't it possible to share the learning results of robots with humans by using professional human skills as robots?



2) E.g.Collaboration: Collaboration between robots and humans is relatively easy to imagine visually, and here will discuss communication methods and settings first. Second, it sounds difficult when it comes to collaboration between algorithms and humans, but that's what you do in the search window.


❻Topic from Panelist

1)Mr. Miyaguchi is in charge of developing construction robots in the research department of a principal general constractor, in Japan. He has experienced the development of construction machinery robots (boom?) in the 1980s and 1990s. As of 30 years later, he is shifting his interest to thinking about how to spread robots to society by analysing in detail factors that have not yet taken root in the field.



2)Mr. Kawakami gave a topic, training by VR software for welding, which is one of the typical complex tasks in the construction industry, and an example of actually improving his skills.


3)Dr. Xie introduced several projects that combine machines as an extension of the body and AI as an extension of the brain. Many guidance systems use AI, and I realised that the concept of successfully affording human actions is the key. Especially for projects that expand the possibilities of the body by attaching relatively small machines such as tails and artificial arms to the human body (the shape may be close to the human arm), the control that blends into the body movement without discomfort is It seems that it is not straightforward. I felt that the topic of this discussion, "learning robots," might be involved.


4)Based on the above paper (case study), we asked Mr. Takasaki, the builder of the palace carpenter, how he had been educating the particularly advanced techniques of the palace carpenter in recent years. I wondered if it was possible to learn by digitising such techniques. The educational policy of the palace carpenter is that it takes a long time to handle the tools, as the phrase "three years of planes, five years of chisel, eight years of marking, and a lifetime of sharpening", and the teaching method will change in the last few years. Is there? I asked the question.


❻Contents of Discussion

①Tools and mechanical devices for enhancing "body function" must have always been next to humans, and with the advancement of technology, their strength and precision have increased. On the other hand, I tried to declare that the concept of tools and algorithms for "strengthening the way of feeling and thinking" is relatively new. However, in reality, the world may be moving forward unexpectedly.

❻議論の内容 / Contents of Discussion

➀「体の機能」を強化するための道具や機械装置は常に人間の隣にあったに違いなく、技術の進歩とともに、その強度や精密さが増してきた。一方、「感じ方、考え方の強化」のためのツールやアルゴリズムという概念は、比較的新しいのではないかと宣言しようとした。 しかしながら、実は世の中は意外と先へ進んでいるのではないか

-The emergence of automatic transmission limited licenses also defines human abilities based on the intervention of machines and control technology, and the laws have already been established. I thought again that it was a phenomenon that had already begun.

- Information technology such as search engines may be said to be a human-computer interaction (HCI) act that has permeated the body and consciousness of the generation of digital natives. It's a dream story a long time ago, such as holding a smartphone in your hand and arriving at a place you've never been to while searching for the best method.



2)The two companies are so closely intertwined that the Cartesian dual classification of "body and consciousness" cannot be made, and original scientific research has already been developed regarding the system "sports". According to researchers in the field, analyzing skilled professional athletes with tracking techniques such as motion capture has some discoveries, but it seems to be a topic in recent years that learning it is a different issue. It is not enough for the player to understand the result of analysis from a third-party perspective, and the first-person perspective on how to coach it and reflect it in the subject action is also the academic discipline of cognitive science. It seems that it is becoming an important issue in the field.


Such discussions are developing in a field called "Braintech" due to the emergence of strong AI in recent years.


❼Brief Conclusion

1) It was possible to share knowledge and skills among computer humans by digitizing through scanning behaviour. There, a new concept of cybernetic learning seemed to be a new concept. However, what can be done in the field is that the first step is to firmly "write down" the implicit thoughts and behaviours.

2)The active palace carpenter has an essential part that cannot be replaced with letters, no matter how accurately he writes it in a book or words. They do not know how to use the method to convey the secret to others. I was reminded that there was also a negative aspect to denying the above declaration. (Auther(Dr.hotta) interpreted that this caution is the same as the above-mentioned cognitive science researchers have stated. Even if the technique can be objectively recorded and described from the perspective of a third party, it is not enough. It requires another step from one's perspective; You need to learn the knack of behaviour from your point of view, to master the ‘real’ skill.)

3) As a first step, it will be easier for humans to interact with it if the robot knows what it is thinking. Isn't such an idea practically possible even with industrial robots? The first step in interaction is to express intentions at a turning point where mass production and the use of robots for automation are transformed into collaboration with humans in Industry 4.0. Opinions were raised, and everyone agreed with the candid views.





The story didn't go as far as the whole concept of digital craft, but it gave us frank and hands-on feedback on learning in a human-machine weaving cybernetic environment that will change.


202204_Responsibility; Dr.Hotta,K