“Most architects tend to see robotics as something very far in the future. Do you think that this is causing them to be left out of the conversation?”
I was recently asked this question during a talk at an architecture program. My response, which I will elaborate upon here, was that automation and robotics already exist within the discipline—but they have manifested differently than what people typically envision. Building information modeling has been steadily automating aspects of architectural design for years. Despite its impact on daily design and documentation tasks, BIM has failed to empower architects with significant control over the building delivery process. The conversation from which architects should not want to be excluded then becomes: How can we more effectively apply automation and robotics technologies that exist today in a transformative manner for project delivery and revenue generation?
The Role of Automation and Robotics in Architecture
To better understand the discipline’s approach to the integration of robotics into practice, architects can reflect on their own education. Architecture students are often introduced to concepts of industrial automation and autonomous robots through the mythos of the “master builder,” a historical precursor to architects today who oversaw not only the design but also the construction of buildings and structures. The implied narrative here—that the profession’s destiny is to reclaim disciplinary domains that became specialized and separated decades, even centuries ago—manifests itself in contemporary discourse as a fixation on the direct control of construction tasks, achieved robotically or with other means of industrial automation and driven by data contained within a digital design model. These machine-readable deliverables are different from traditional drawing sets in that they don’t require a skilled worker to translate design intent into constructed action. Instead, they bypass tradespeople as a means to reclaim “craftsmanship” and centralize building knowledge to, in theory, dictate design intent more directly and deeply into the building delivery process.
This file-to-field approach, challenging even for the few vertically integrated organizations currently capable of its execution, is infeasible to implement in today’s prevailing risk-averse and siloed design-bid-build delivery model. In addition to being categorically misleading—AIA contracts prohibit architects from engaging in “means and methods”—this approach both distracts from and misses the greater potential of robotics in the field of architecture, such as opportunities to expand professional services and revenue models premised on autonomous means of jobsite data collection. Data, it turns out, is both the lifeblood and the inherent limitation of BIM. The challenges of deriving value from BIM more than a decade after its industrywide adoption continue to center around the inability to process and supply data in the right quantities and formats to the right stakeholders at the right time.
Remote Construction Administration and Jobsite Telepresence
Architects neither own the construction site nor operate primarily within it, so they understandably struggle to access the jobsite data they need in order to augment their design models with as-built conditions. Hamstrung construction administration (CA) practices disconnect designers from construction site professionals: Typically, only a small number of architects visit the job site periodically, while the rest of the office stays fixed to stationary computers.
While any communication lapse has undesirable outcomes, the true danger of a siloed design culture is its contribution to misguided notions of how information flows. Model-driven delivery methods are premised on the notion that the intelligence required for the delivery of a building is holistically contained within the design model, when in fact much of the intelligence required for the construction of a building comes from the tradespeople on site. While intended to exert design intent deeper into project delivery, a linear delivery process actually stymies opportunities for productive collaboration between architects and tradespeople. When design intent is presented as a set of fixed information, without the ability to accommodate feedback or adapt to site conditions, it frames CA as an exclusively “quality-control” process rather than as a cooperative or creative one.
This decentralization of knowledge is not an impediment to building delivery, but rather an inviolable requirement. Successful delivery depends on the extensive experience in the field; when decentralized intelligence isn’t accommodated by delivery technology, upstream communication has to swim against the current of a linear data flow. This effort is costly and contributes to the inadequacy of as-built information in the model, hampering effective design documentation and complicating owner handoff.
From Linear Design–Delivery to a Field Feedback Loop
How do architects reduce upstream data friction to drive down the cost of an effective data feedback loop with the field? They can start with a more robustly connected job site. Multiple parallel developments over the past decade—mobile devices in every pocket, 5G and LTE edge networking, cheap and portable sensors, and well-trained artificial intelligence—have made automating not only the remote capture of unprecedented amounts of site data, but also its upload to cloud-based project tracking platforms for analysis easier than ever. The need for this data infrastructure has been exacerbated by COVID-19 as personnel density restrictions have increased the amount of people who require jobsite telepresence and remote data access. What this data infrastructure lacks, but has been collectively enabled by recent technological developments, is autonomous sensor mobility in human-purposed environments.
Enter agile mobile robotics, the missing piece in an autonomous field-data tech stack that overcomes current limitations of static sensors and aerial drones for interior capture in order to perform a precise, unsupervised repetition of tasks at scale. (Disclosure: I work at Boston Dynamics, which offers this technology with its Spot robot. Prior to joining Boston Dynamics, I was a construction industry researcher and had the opportunity to test Spot against competitive solutions.) This new breed of terrestrial robot, legged instead of wheeled or tracked, can walk over and around obstacles and ascend and descend stairs. It can dexterously navigate active construction environments while carrying sensor hardware and onboard software that captures data essential to project management, such as laser scanners, 360-degree cameras, thermal imaging devices, and environmental sensors. I detail a selection of use cases below.
As-builts
A truly accurate as-built document—a precise digital counterpart to what was actually built rather than a design intent model—offers a particularly compelling opportunity for firms to leverage their expert services later in the delivery process (see “Handoff” below) and, frankly, to learn: All too often architects, in a fugue of deadline-induced haste, contractual impotence, slim margins, and the manifold layers of real estate market abstraction from end tenants, move from project to project with no real sense of whether their product works. As a result, they gain no insight to apply to the next product iteration.
Industrialized Construction
As-builts are inherently limited in that they represent a static end state, but augmenting the building information model with some resolution of time—think hertz but as units of construction information—enable these models to continuously incorporate jobsite data such that designers can manage design intent relative to the realities of the jobsite. As prefabricated practices become critical drivers of construction efficiency, a continuously up-to-date model to ensure the dimensional compatibility between prefab building components and the building structure will become essential. This is particularly true for commercial renovations, but even ground-up projects deviate dimensionally from the design model due to unanticipated realities.
Handoff
Over the full life cycle of a building, as-builts offer value to not just architects, but also building owners. They can enable effective post-occupancy analyses and the creation of additional services around real estate and asset management. Owners often find it difficult to reconcile the digital representations of their physical assets in BIM with their physical assets because the industry has assumed that 3D geometry and manually input metadata are intelligible. In fact, asset data within as-built models might be better accomplished with reality capture data types such as point clouds and photospheres, which can be processed for semantic interpretation using computer vision and other learned models (such as Smartvid or Avvir), or with RFID or BLE tag scans that can be correlated with digital asset via cloud databases (such as Orbcomm).
When this data is autonomously captured by machines, it has the added advantage of a built-in digital log that establishes trust among stakeholders. Dusty Robotics’ construction layout robot is driven by data from BIM and/or digital shop drawings, which is already hugely valuable in and of itself. Even more exciting, the robots could potentially log field adjustments in a BIM-readable format for seamless as-built feedback into the model.
Transforming Talk of Data into Action
While the value of automated data capture on job sites is clear, the stakeholder who is ultimately responsible for funding and controlling capture operations is not. Data capture is not explicitly within a subcontracted trade’s scope; while a general contractor would obviously benefit from it, no clear precedent for rigorous jobsite reality capture at scale exists.
Architects do not have the authority to access construction sites at-will in order to manage equipment—and it’s hard to imagine a design firm owning a fleet of jobsite robots today. Then again, the very existence of these robots in construction environments was hard to image just a few years ago, and the establishment of the new services described above could be passed along to the client who benefits. Furthermore, construction robots are networked, remotely accessible, and thus easily shareable between stakeholders. They can be used across multiple projects and/or beyond construction into the operation of a property, justifying the cost of the initial investment with deployments over the full building life cycle.
Broader yet, architecture practices could pool resources through the collaborative investment in more entrepreneurial-minded architectural ventures, such as the recent forming of Spatial Syndicate. As Base2 Corp. founder and CEO John Manoochehri put it when describing the AEC startup and AI tech exchange Hypar, what architecture needs right now is to foster a “culture of collective improvement.”
Architects have always talked a big game about data, despite their limited access to it. Technology is enabling a level of access to field data that architects have never before experienced. An architect’s ability to consistently access a site, whether it be physically or remotely via agile mobile robots, creates a simultaneity to design and construction practices. This enables a positive feedback loop that encourages productive dialogue between stakeholders and the continuous collection of job site data for incorporation into a design model, finally delivering on the value propositions of BIM. With remote work remaining the norm for the foreseeable future, the solutions discussed above can transform the talk of data into meaningful action, and thus transform the entire building industry.
The views and conclusions from this author are not necessarily those of ARCHITECT magazine nor of The American Institute of Architects.
