Goal-driven design in a bio-inspired system

Abstract

Top-down systems tend towards optimising a design space by splitting it into exploitable parts. Bottom-up frameworks seek to explore the design space by working with less constrained parts that are then used to form solutions. Both approaches are useful within design thinking and a balance between exploration and exploitation is to be found. This thesis seeks to address a gap around using explorative algorithms and introduces new concepts to exploit a design space, by increasing communication between the design process and the goal. The approach developed can be identified as a type of generative design.

Framing the direction of work, first the many system goal definitions in literature are consolidated into a formal statement which forms the foundation for this new approach . This is then used to create a framework for implementing the goal using a system comprised of four parts - the goal, objectives, measurables called logics and actionables, the components which do the work.

The second major question addressed focuses on how to analyse and evaluate parameters within an exploratory environment where the model may not be complete. Two new concepts are developed and introduced here: restriction – a decision point, and misfit – a set of measurables. Combined they provide the design system with a quantifiable reference of the goal, helping to influence growth direction within an exploratory algorithm.
The final implementation is demonstrated using a bio-inspired design system and the novel methodology of gene regulation. Here the growth isn’t changed directly by analysis but influenced through design strategies which explore the design space. These strategies are implemented and adapted through the misfit function which is derived from the goal. This helps to ensure that the design follows unique paths, leaving room for emergence, and yet is still heading towards the goal, producing a feasible solution.

Date of Award	Jul 2023
Original language	English
Awarding Institution	Queen's University Belfast
Sponsors	Engineering & Physical Sciences Research Council
Supervisor	Declan Nolan (Supervisor), Mark Price (Supervisor) & Wei Zhang (Supervisor)