Arms Race

An arms race is on in the worlds of com­pu­ta­tion and archi­tec­tur­al fab­ri­ca­tion research. Robots with increas­ing­ly large, fast, and pow­er­ful capa­bil­i­ties are avail­able and can pro­duce out­puts with mil­i­tary-grade pre­ci­sion. The assump­tion is that, through the use of these advanced tools, archi­tects will also advance the pro­duc­tion of out­puts, but can these tools be devel­oped with tra­di­tion­al forms of human engage­ment still in mind? Robots are not par­tic­u­lar­ly adap­tive. They do not inte­grate changes with ease — at least, not yet. Humans, on the oth­er hand, exhib­it great capac­i­ty for adap­ta­tion but lack the pre­ci­sion of robots. How could pre­ci­sion and adap­ta­tion be com­bined in archi­tec­ture, specif­i­cal­ly with­in the con­text of Japan, where I live and work and where imper­fec­tions are embraced as part of an ide­al form?

Explor­ing that ques­tion was one of the aims behind the STIK (Smart Tool Inte­grat­ed Kon­struc­tion) Pavil­ion project.

Through year­ly lab­o­ra­to­ry projects, stu­dents in Advanced Design Stud­ies at the Uni­ver­si­ty of Tokyo (T_ADS), where I teach archi­tec­ture, have the oppor­tu­ni­ty to explore non-tra­di­tion­al uses of oth­er­wise tra­di­tion­al mate­ri­als and approach­es. Those projects pro­vide stu­dents with unique, new chal­lenges in terms of fab­ri­ca­tion, con­struc­tion, and com­pu­ta­tion while encour­ag­ing lim­its to be pushed.

For the STIK project, we focused col­lec­tive­ly on the devel­op­ment of a scaled-up ver­sion of a 3‑D print­er that would uti­lize chop­sticks. The lat­ter mate­r­i­al was cho­sen because, in large quan­ti­ties, it is capa­ble of behav­ing like par­ti­cles, can be aggre­gat­ed, and can be solid­i­fied using adhe­sives. Addi­tion­al­ly, its scale as an aggre­gate is more con­ducive to pro­duc­tion of a build­ing, unlike the small­er-scale mate­ri­als used in con­ven­tion­al 3‑D print­ers. To accom­mo­date our mate­r­i­al, we also want­ed to make a fab­ri­ca­tion tool capa­ble of aug­ment­ing human capa­bil­i­ties through enhanced net­work­ing, which we called a net­work tool.

The net­work tool was devel­oped in two parts: a visu­al guid­ing sys­tem and a chop­stick-dis­pens­ing machine. The visu­al guid­ing sys­tem assists its oper­a­tors via a pro­jec­tion map­ping sys­tem and a scan­ning sys­tem using a video pro­jec­tor and a Kinect; a 3D motion sens­ing and scan­ning device devel­oped as part of Microsoft’s video game con­sole. The guid­ing sys­tem facil­i­tat­ed oper­a­tors’ knowl­edge of where, how, and when to oper­ate the machine, thus achiev­ing an adap­tive, real-time, dis­trib­uted fab­ri­ca­tion process. This com­bi­na­tion of man­u­al labor and com­pu­ta­tion­al analy­sis required not only the devel­op­ment of hard­ware but also soft­ware and human-machine inter­faces. The dis­pens­ing machine, devel­oped specif­i­cal­ly to deposit a defined amount of chop­sticks at a spe­cif­ic loca­tion, has char­ac­ter­is­tics that may be applic­a­ble to oth­er scaled-up 3‑D print­ing process­es. The key dif­fer­ence between our approach and more con­ven­tion­al print­ing process­es is that the dis­penser is car­ried by its oper­a­tor as a mobile, hand­held device. Dis­penser devel­op­ment aimed to address the fol­low­ing goals:

1) Production/​assembly of a form with­out the use of a phys­i­cal formwork.

2) Facil­i­ta­tion of con­struc­tion process­es in real-time; out­put is not a pre­de­ter­mined tar­get, but rather is con­stant­ly being adjust­ed with a feed­back sys­tem con­nect­ing scan­ning sys­tem, struc­tur­al analy­sis, and dis­pens­ing process.

3) All ​“print­ing” devices are net­worked such that the process can be mon­i­tored on both the local and glob­al scales.

Our dis­penser facil­i­tat­ed con­struc­tion process­es by act­ing as a link between the phys­i­cal form and the dig­i­tal mod­el. Mul­ti­ple net­worked dis­pensers used at mul­ti­ple loca­tions across a con­struc­tion site can con­tribute to the cre­ation of a sin­gle out­put in a swarm-like process, and the progress can be checked through­out assem­bly. The sys­tem is also capa­ble of scal­ing, some­thing which is lim­it­ed in typ­i­cal 3‑D print­ers because of their size.

Chop­stick Dis­penser Machine Devel­op­ment

Our machine evolved as the project require­ments and ambi­tions evolved. The first iter­a­tion relied entire­ly on the use of man­pow­er. A hand-crank turned the mech­a­nisms with­in the machine and dropped chop­sticks. This evolved into a ver­sion with a DC motor, which sat­is­fied a require­ment for con­sis­tent chop­stick drop­ping speed. In ini­tial ver­sions of machine devel­op­ment, glue was not inte­grat­ed; work­ers added glue by hand with a brush as out­puts were ​“print­ed” with chop­sticks. As machine devel­op­ment pro­gressed, a mod­i­fied glue feed­er was inte­grat­ed into the body of the machine, which helped pre­vent the glue from clog­ging the dis­pens­ing process while also ensur­ing chop­sticks were coat­ed before drop­ping onto the con­struc­tion site.

Space was includ­ed on the machine for an Arduino board, a low-cost, sin­gle-board micro­con­troller that served to reg­u­late drop­ping speed and the amount of glue as well as to indi­cate the rela­tion­ship of the machine to oth­er machines in the con­struc­tion process — and, thus, to pro­duc­tion over­all. Full inte­gra­tion of the Arduino sys­tem could allow inde­pen­dent work at mul­ti­ple sites on a project with mul­ti­ple teams of work­ers. Through its capac­i­ty for adap­ta­tion and the inte­gra­tion of new infor­ma­tion, the sys­tem can pro­duce out­puts using dig­i­tal form­work, smart tech­nolo­gies, and net­worked communications.

The STIK Pavil­ion was not a pure­ly aca­d­e­m­ic project; it was a col­lab­o­ra­tive effort with Shimizu Cor­po­ra­tion, a Japan­ese con­struc­tion com­pa­ny. Two of their pri­ma­ry con­cerns were:

1. Pro­vid­ing a method of assist­ing less-skilled labor­ers as num­bers of skilled work­ers decline, and

2. Cost-effec­tive on-site management.

Their approach to con­struc­tion includes off-site fab­ri­ca­tion, which allows all indi­vid­ual parts to be fab­ri­cat­ed through indus­tri­al process­es, brought to the site, and assem­bled through human involve­ment and effi­cient assem­bly approaches.

The item miss­ing in their approach was the inte­gra­tion of a net­worked tool for con­struc­tion crews, which we pro­vid­ed. Over time, tools have devel­oped from sim­ple hand tools to pow­er tools.

Net­worked tools are the log­i­cal next devel­op­ment in that progression.

Fig­ure 1. Pro­gres­sion of Tools

Net­worked tools allow oper­a­tors to aug­ment their under­stand­ing of con­struc­tion process­es in real time.

The STIK Pavil­ion project became an oppor­tu­ni­ty to test that capac­i­ty. The dis­penser we devel­oped assists and aug­ments human involve­ment at the con­struc­tion site and dur­ing the con­struc­tion process. Over­all, the ambi­tion of the project extends beyond mak­ing a pavil­ion with chop­sticks to explor­ing the trend of 3‑D print­ing tech­nol­o­gy and exam­in­ing how sim­i­lar print­ing process­es can pro­duce archi­tec­ture at full-scale at the con­struc­tion site. In this sys­tem, humans become agents who can be aug­ment­ed and net­worked through the use of smart tool devices. Human arms and robot­ic arms can work togeth­er to achieve results that would be impos­si­ble with­out collaboration.

Fig­ure 2. Evo­lu­tion of Adap­tive Design Strategies

In past work at the Archi­tec­tur­al Asso­ci­a­tion Design Research Lab in Lon­don, I focused on the sim­u­lat­ed rela­tion­ship between mate­r­i­al com­pu­ta­tion and dig­i­tal com­pu­ta­tion to cre­ate adap­tive design pro­pos­als (Fig­ure 2, Adap­tive Design Process). Sim­u­la­tions would exam­ine a mate­r­i­al behav­ior and then sim­u­late that behav­ior in the dig­i­tal envi­ron­ment before attempt­ing to pro­duce that behav­ior with ana­log mod­els. Sub­jec­tiv­i­ty and ran­dom­iza­tion (pro­duced by humans) was min­i­mized as much as possible.

In the present approach (Fig­ure 2, Adap­tive Con­struc­tion Process), how­ev­er, the human aspect is not only includ­ed in the fab­ri­ca­tion process; it is also embraced. Incon­sis­ten­cies intro­duced to the project by humans are nav­i­gat­ed with net­worked tools, which facil­i­tate ongo­ing mate­r­i­al and dig­i­tal com­pu­ta­tions. Such tools allow for pro­duc­tion of unique out­puts that inte­grate human ran­dom­ness, spec­i­fied forms, and dig­i­tal com­pu­ta­tion. The forms pro­duced through this sys­tem are not com­plete­ly pre-defined through com­pu­ta­tion­al sim­u­la­tion. Although the forms can be sim­u­lat­ed to a degree in a com­pu­ta­tion­al envi­ron­ment, they can­not be repro­duced accu­rate­ly due to lim­i­ta­tions relat­ing to human and mate­r­i­al behav­ior. The net­worked tool facil­i­tates a solu­tion for this issue, and ulti­mate­ly achieves what Amer­i­can econ­o­mist Leo Cherne hint­ed at over a quar­ter-cen­tu­ry ago, ​“The com­put­er is incred­i­bly fast, accu­rate, and stu­pid. Man is unbe­liev­ably slow, inac­cu­rate, and bril­liant. The mar­riage of the two is a force beyond cal­cu­la­tion.”¹