Dry Stacking Strategies for Autonomous Construction
Thangavelu, Vivekanandhan S.
MetadataShow full item record
In scenarios that lack modern construction infrastructure, building with found objects might be the only process available. For example, in remote environments such as extraterrestrial construction where the cost of transport is limiting, or in disaster areas where temporary structures need to be built quickly and possibly without a functioning supply infrastructure. The specific question addressed in this thesis is how to enable automated construction with found objects without the use of a binding component. The advantage of such an approach is that in some environments, materials can be found in abundance and are essentially free in terms of production effort and transport cost. The thesis describes a method for automatically building structures from stacked, irregularly shaped objects. A simplified 2D model for the problem of building dry stacked structures (i.e. no mortar) from found stones is presented and analyzed. Currently such structures need to be built by skilled masons. No practical methods for automating the assembly planning process are known. The problem is challenging since each assembly action can be drawn from a continuous space poses for an object and several local geometric and physical considerations strongly affect the overall stability. We show that structures that are built following a stacking order for perfect bricks can accommodate a limited amount of irregularity, however, their performance degrades quickly when objects deviate from their ideal shape. We introduce a quantitative analysis of how the planning strategy is closely coupled with the object shape. We present a strategy for stacking irregular shapes that first considers geometric and physical constraints to find a small set of feasible actions and then further refine this set by using heuristics gathered from instructional literature for masons. The proposed method of choosing assembly actions enables construction with objects that contain a significant amount of variation.