Paper title: |
Using generative representations for structural design |
Authors: |
Yu Zhang, Alan Kwan, John Miles |
Summary: |
Work in recent years has shown that topological reasoning with search algorithms using traditional rep-resentations such as parameters, ground structures, voxels, etc is very limiting. Each type of representation is only to be suitable for a limited number of topologies. This is restrictive because there are many problems where the topology of the solution is unknown except in the most general terms or there are competing topologies which are suitable for solv-ing a given problem. Hence, at best, choosing a representation technique can be difficult and at worst it can restrict the search so that a full examination of the problem is not possible. Also, as the available computational power increases and the technology of search algorithms is enhanced, the topologies being reasoned about become ever more complex and so the representations within the algorithms can become cumbersome. A possible solution to these difficulties is the use of generative geometries where the object is represented by a set of rules which describe how to create the object. These can, when correctly implemented, give a compact representation and one which can be handled within typical search algorithms like for example genetic algorithms. This paper looks at the use of L-systems. They are being applied to beam design problems although this paper focuses on the representation. As will be shown in the paper, although the representation has some attractions, there are also some difficulties with the implementation and especially with en-forcing constraints. The paper describes work which is in progress rather than a completed project. |
Type: |
|
Year of publication: |
2007 |
Keywords: |
generative representation, evolutionary computation, structures, search algorithms |
Series: |
w78:2007 |
ISSN: |
2706-6568 |
Download paper: |
/pdfs/w78-2007-101-131-Zhang.pdf |
Citation: |
Yu Zhang, Alan Kwan, John Miles (2007).
Using generative representations for structural design. 663 (ISSN: 2706-6568),
http://itc.scix.net/paper/w78_2007_108
|