








The Inner Forms project explores the intersection of robotic fabrication, functionally graded timber structures, and machine learning-driven material optimization. Inspired by Frei Otto’s interpenetrating shell systems, the research investigates parametric form generation, laminated timber construction, and adaptive milling techniques to enhance both structural performance and fabrication efficiency.
At the core of this research is the development of functionally graded timber structures through hybrid cross-lamination strategies, allowing for controlled variations in density and structural behavior. Using a combination of Douglas Fir, Ash, and Walnut, the study evaluates material properties and their influence on fabrication methods. Robotic milling and precision joinery techniques enable the seamless assembly of complex wooden components, ensuring structural integrity while maintaining material efficiency.







A key innovation in this project is the integration of machine learning for real-time material detection and robotic adaptation. The system incorporates both a YOLO-based wood species detection model and a sound-based detection model that identifies wood types during milling. The sound-based model enables autonomous adjustments to milling speed, ensuring optimal cutting strategies based on the unique density and grain structure of each wood species. This dual-sensor approach enhances efficiency, precision, and adaptability in robotic timber processing.
The fabrication process follows a multi-step workflow, beginning with wood delamination, where different layers of timber are separated and reassembled using a barrel construction method to introduce graded material properties. Advanced non-planar milling paths are then employed, allowing for greater geometric freedom and smoother transitions between material layers. This method not only improves surface quality but also optimizes structural performance by strategically varying the material density across the form.
The project culminates in a real-time interaction system, where computational design, robotic control, and AI-based material recognition converge to create a fully automated, intelligent timber fabrication workflow. By bridging parametric design with material intelligence, this research presents new possibilities for functionally graded timber structures in architecture and fabrication.
















Master of Science in Design: Robotics and Autonomous Systems (MSD-RAS)
Weiztman School of Design, University of Pennsylvania
ARCH 8011 Material Agencies II_ Fall Semester
Instructors: Alicia Nahmad Vazquez & Patrick Danahy
TAs: Mahsa Masalegoo & Jean-Nicola Dackiw & Soroush Garivani
MSD-RAS Program Director: Robert Stuart-Smith
ARI Robotics Lab Operators: Nicholas Sideropoulos & Shunta Moriuchi
Team: BurcuGocen, BeikelRivas, QingyangXu, ZitongRen