Project Fractal

Autodesk’s Project Fractal, which began development in November 2015 and entered public alpha testing in July 2016, is created as an open system applicable to multiple design problems, employing scalable cloud computation to generate proposals while facilitating rapid evaluation of those solutions. Highly-performing design candidates assessed by custom evaluators can be selected for further exploration and elaboration. While such tools have been developed and explored in a number of academic institutions and professional firms for decades, the availability of cloud computing now affords the opportunity for the AEC industry and academia to more widely accelerate their computational delivery of professional expertise and the advancement building design, construction, and operations in conjunction with academic investigations.

Where design and construction firms may invest considerably in the study of a few alternative designs or construction approaches, Project Fractal’s use of scalable computation allows the designer to range more widely through possible solutions generatively produced, prompting further exploration along avenues previously deemed impractical due to aggressive schedules and fee constraints. By employing generative and analytical techniques, designers and builders can quickly evaluate an extensive array of alternatives for a critical building decision.

Autodesk’s collaboration with Perkins+Will from January through August of 2016 resulted in the prototype Space Plan Generator graph in Autodesk Dynamo that accounts for a number of key goals, requirements, and constraints applicable to a hospital bed tower project. Incorporated evaluation functions indicate to the designer the relative success of each proposal linearly generated from manipulations of input factors. Submitting the graph to Project Fractal creates multiple plan alternatives generated from varying combinations of key parameters for comparison by the designer both visually and through identified metrics.

Project Fractal’s precursor, Project Akaba, demonstrated a concept of generative space planning but was limited by compiled code that created the proposals. A few opaque algorithms could process a building program in different ways to attempt to solve a space planning problem, but if someone outside Autodesk had a better idea, there was no way to instantiate that idea to be expressed in the Akaba environment.

The primary motivation for using Dynamo as the platform and Dynamo Studio as an editing environment for Project Fractal is the impossibility of centrally capturing all relevant building expertise. Revit attempts to convey a minimal complement of building component structure and interaction behaviors, but even those affordances can be limiting in some situations because the rules conceived for buildings modeled within Revit do not obtain for every building, everywhere, even if they’re perfectly fine for many buildings in many places.

Like Dynamo, Project Fractal is conceived as a system to help capture and deliver expertise as recorded with visual programming. We’re working on a platform to use graphs with designated inputs and outputs to explore a design space, visually at first and gradually with the addition of relevant analysis and automated goal convergence. The further we pursue such ambitions, the more we find problems that have, at best, a few plausible approaches but no definitive answers for describing goals and constraints in an accessible and portable manner. For now, we’re creating and trying out a few sample graphs in the system.

ANTHONY HAUCK

Autodesk Inc.

Anthony Hauck has been involved in architecture, engineering, construction, and technology for more than 30 years. As an architect, millwork project manager, software developer, and IT Director, he has always looked to technology to help solve issues facing the building industry. He joined the Autodesk Revit team in 2007, holding a succession of product management positions in the group until joining Autodesk AEC Generative Design in 2015 as its Director of Product Strategy, where he is responsible for helping define the next generation of building software products and services for the AEC industry.

Translate »