Closed Arrow Icon Topic

BUILDING INSTINCTS - EVOLVING NEURAL CONTROLLERS FOR CREATIVE MACHINES

 

The integration of digital production into architecture carries the potential to unite the realms of design and construction in a way that questions many of the well established renaissance notions of the discipline’s competences and responsibilities. If the theorist Mario Carpo and the proponents of robotic fabrication are to be believed, we are already in the midst of this turnaround. Recent technological advances make it possible to incorporate real-time data into digital production processes, which reduces the necessity to define the exact sequence of events in advance. Thus, the potential roles of the machines involved are extended - from mere automats, aiming at the highest possible accuracy in the reproduction of a previously developed model, to active players, whose place in the design process has yet to be found. It also enables them to leave their well-defined surroundings of labs and production lines for more uncertain and dynamic environments, without the need for direct control by a human operator. Along with this prospect of higher autonomy and mobility, questions of design turn into questions of machine behaviour: How, in what form and to what extent can it be defined so that its material manifestation reflects design intent? In relation to this question, the thesis is investigating if a method that is used for the development of robot control systems and in artificial-life research can be applied to autonomous building. Its integration in a design process, the relevant technical fundamentals and developments are discussed. It is implemented as a software framework using a custom plugin for the powerful robot simulator v-rep. In addition, a conceptual robot design for the assembly of lightweight rod structures is presented, and its sensorimotor and control system are implemented in a simulated building process.

 

Closed Arrow Icon Supervisor