Habitat Prospecteur (Master Thesis, ENSAPM 2012)
For this thesis project, I decided to address and merge the architectural topics that influenced me the most during my apprenticeship:
- geometry and structural optimization
- programming and simulation
- manufacturing and assembly processes
The goal of this project was to develop, fabricate and build a robotized micro-house prototype answering to the following constraints:
- A total budget of 1000€
- A 580mm robot arm with a playload of 3Kg
- A simple car to move the parts
- No extra material to produce tests / verify robot calibration
- A 2 weeks prefabrication period
- A 2 days assembly period requiring almost no equipment (glue gun)
- A fully optimized scripted 3D model in order to allow complex simulations and automated robot programming.
The design in itself used the ability of the robot to produce precise foam parts using a hotwire cutter, to create a structural foam shell that could be coated in order to resist to tension stresses or used to cast concrete, thus reducing the cost and weight of the building structure. PVC parts (from tubes) where maintaining the foam parts together, and hot glue was sprayed to joint them together.
The design process included several optimization steps:
- A shell discretization optimization in order to minimize the amount of parts while respecting the dimensions and topological limitations inherited from the robot properties.
- A Shell refinement algorithm that used isocurvature domains of the surface to compute and minimize the boundig boxes of the different parts.
- A complex 3D nesting algorithm that allowed to nest the parts using a system of priority depending on the available foam blocks dimensions, the cutter precision, the thickness of the glue, and the possible orientations of the block in front of the robot.
- 160 unique foam components (approx. 500*600*80mm, curved on at least 5 of their 6 faces). All the blocks were extracted from 4 1200*1200*600mm and 4 1000*1000*500mm 30Kg/m3 foam blocks .
- The cutting was done using a custom 670*300mm hot wire cutter (bigger than the robot itself… but that allowed to reduce the amount of parts by 30%).
- Approximately 120 hours of cutting (so approx. 45min/part including positionning of the primary block, pre-cutting, positioning of the raw part, final cutting, dimensions check and naming).
- 2 definitions of approx. 2600 components each: one for the geometry generation, optimization and 3D nesting (takes 20.7s to compute a whole micro-house and all its parts); and another one for tool path extraction, kinematics solving and robot instructions generation (takes 600ms to compute and simulate the 1800 lines of code for each part to cut).
- Assembly in less than 30 hours (including the timber framework) by 3-4 untrained students with a high-pressure glue gun, a compressor, an electric screwdriver and a stepladder.
- All the parts, including the timber framework and the assembly equipment have been moved in two times using a Seat Alhambra.
- Prototype dimensions: Length = 5000mm; Width = 2000mm; Height = 3500mm.
- The prototype was not finished with a coating, to allow its disassembly after the presentation. It is know waiting for it in the garden…
Below, the set of explanatory graphical documents, pictures and videos of the simulation and fabrication.
Toolpath generation and simulation steps
Morphogenesis, optimization, discretization of the shell; 3D nesting of the shell parts.
Extraction of the toolpaths from the parts geometry for both the raw bounding box extraction and the final 6-faces cutting.
Simulation of the toolpath, translaton in robot native language and dynamc upload to the controller using the HAL plugin.