While many digital fabrication processes in architecture focus on using standardised prefabricated materials, NCCR Digital Fabrication researchers Martin Wermelinger, Fadri Furrer and Hironori Yoshida (former) aim to use unprocessed materials found on-site to create new structures. An interdisciplinary research team from the Autonomous Systems Lab (ASL), the Robotic Systems Lab (RSL), and Gramazio Kohler Research (GKR), from ETH Zurich combines knowledge in the areas of object detection, simulated construction, robotic manipulation, and structural planning to address this novel idea. As a proof of concept, the team has now successfully achieved autonomous robotic stacking of irregular stones without the use of mortars or adhesives.

In the last decades, great advances have been made in the robotic assembly and processing of a wide variety of manufactured goods. However, until now these technologies have not translated well into the construction sector.

Potential of Robots to Enhance Sustainability in Construction
Robots are ideal machines to undertake mundane tasks of repetitive nature where physical force and stamina are required. Additionally, they are especially practical for environments where worker safety is of issue. When applied in construction, robots have strong potential to enable innovative building processes. However, they have to be equipped with additional sensors, feedback control, and the right programming to deal with the uncertainties and irregularities of such environments. With this enhanced skills, they can provide new opportunities for sustainability in construction.

Recycling Found Objects as Building Blocks
The wider objective of this research is to develop methods for the automation of a complete construction workflow. By combining robots with sensors and responsive algorithmic control systems, the technologies can be put to work in ways that are both autonomous and responsive to their environment. More specifically, the project aims to detect and scan single irregular objects in cluttered scenes and to be able to compute, simulate and control their robust interactions, all with the goal of using the re-cycled objects as “building blocks”. In order to achieve this, the object forms must be captured and modelled, such that the data is usable in both physics simulations and geometry matching algorithms.

Autonomous Building of Vertical Stone Stacks
A scale prototype system which is able to build vertical stacks of small (10cm scale) stones represents the current state of research: after detecting objects (stones) in an available work field, possible placement configurations for individual stones are computed and then evaluated in physics simulation using a cost function with the goal of identifying stable stacking compositions. The “next best” stone for assembly is then selected by the robot with an appropriate grasping configuration and after planning a collision free motion trajectory the desired placement is executed (Figure 1 and embedded video).

Figure 1: The “next best” placement configuration of a stone is computed and evaluated in a physics simulation (left) and executed by the robot arm (right).

Stability of each stone and the resulting construction is verified by cross referencing the computed against the “as built” condition by using the visual and force-torque feedback sensors within the robot to measuring possible movement. Stacked configurations are the result of explicit design goals such as target shapes, structural parameters, or the in-filling of pre-defined space – all while using available “found” objects.

Further Steps of Research and Possible Applications
In the current proof of concept demonstrator, the stones being used are pre-scanned, enabling the system to recognize the modelled geometries using an RGB-D camera. In a next major step, researchers aim to develop automated processes for the extraction of 3D geometries and to directly create the digital models from the scanned objects. Since online modelling generally increases the uncertainties in the whole construction workflow, new methods employing compliant actuation and force feedback will be developed in order to be able to robustly control the interaction with objects.

All in all, this collaborative research project will open up new possibilities on different architectural scales. Possible future applications are the reutilisation of construction waste material directly on site, reconstructions after disaster events or the autonomous construction of dry stacks and river beds taking desired shape and structural parameters as inputs.

Figure 2: Today, demolition material on a construction site is either brought to a disposal place or has to be (manually) processed to be reused. Creating new structures with unprocessed material directly on-site can have a large impact on the sustainability of the construction process.