A Multi-Agent System for Facade Design: A design methodology for Design Exploration, Analysis and Simulated Robotic Fabrication

Abstract

For contemporary design practices, there still remains a disconnect between design tools used for early stage design exploration and performance analysis, and those used for fabrication and construction of complex tectonic architectural systems. The research brings forward downstream fabrication constraints into the upstream design exploration and design decision making. This paper addresses the issues of developing an integrated digital design work-flow and details a research framework for the incorporation of environmental performance into a robotic fabrication for early stage design exploration and generation of intricate and complex alternative façade designs. The method allows the user to import a design surface, define design parameters, set a number of environmental performance objectives, and then simulate and select a robotic construction strategy. Based on these inputs, design alternatives are generated and evaluated in terms of their performance criteria in consideration of their robotically simulated constructability. In order to validate the proposed framework, an experimental case study of office building façade designs that are generatively created from a multi-agent system for design methodology is design explored and evaluated. Initial results define a heuristic function for improving simulated robotic constructability and illustrate the functionality of our prototype. Project limitations and future research steps are then discussed. 13 PROCEDURAL DESIGN led to the reconsideration of the architect as both a contemporary master-builder and a digital toolmaker (Tamke and Thomsen 2009). This further highlights the need for more integrated approaches for the AEC industry, and an urgency to develop new design methodologies that support a more holistic approach towards the adoption of robotics, not just as tools, but as participants in the multi-objective, complex, and highly synthetic design decision-making process. Despite the numerous advancements, there remains a gap in intuitive workflows for designers to incorporate simulation results as drivers and constraints early in the design process (Kilian 2006). In addition, disparate software is used for different design phases-design generation, simulation, manufacturing, and construction planning-and consequently continuous information exchange among the AEC disciplines still remains a challenge (Scheurer 2007; Schwinn and Menges 2015). Most critically, current design methods and computational tools remain limited in their foresight, and in registering design parameters and constraints from upstream through to downstream processes. Generally, these methods do not consider assembly and construction constraints, nor do they account for the adaptation of projects into local conditions and the reality of real world analogue noise. Currently, there are few robust and efficient programming strategies for controlling multiple robots in custom-ized semi-automated construction conditions. Challenges for such programming tasks include: a) a constantly changing environment (construction sites); b) much smaller production volumes (buildings are one off products); and c) a much larger range of required tasks (Bechthold 2010). Our conjecture is that the next steps in digital design and robotic fabrication are to develop methodologies that consider computers and our robotics as collaborative partners in the design process, as having agency and the capacity to register contextual conditions, environmental analyses , and construction constraints in order to provide architects with design alternatives that fulfill complexly coupled criteria, such as environmental performance, structural efficiencies, and fabrication constraints (Shea, Aish, and Gourtovaia 2005). This research presents our prototyping of an integrated design methodology that allows for validated simulations of the robotic assembly sequence to be integrated into the early phases of the design process. As a first step, we focus on coupling environmental performance simulations and robotic construction simulations for informing and optimizing a generative design process based on our multi-agent systems (MAS) for design work. We validate our framework by applying it to building envelopes (i.e., facades) of a set of commercial buildings representative of varying geometric forms. Building envelopes are among the most complex architectural components, combining aesthetic,