is Íslenska en English

Lokaverkefni (Meistara)

Háskólinn í Reykjavík > Tæknisvið / School of Technology > MSc Tölvunarfræðideild / Department of Computer Science >

Vinsamlegast notið þetta auðkenni þegar þið vitnið til verksins eða tengið í það: https://hdl.handle.net/1946/46267

Titill: 
  • Titill er á ensku A foundation for autonomous conceptual engineering design
Námsstig: 
  • Meistara
Höfundur: 
Útdráttur: 
  • Útdráttur er á ensku

    The engineering method is one of the most powerful problem-solving tools humanity has ever developed. It allows a designer to begin with a high-level conceptual understanding of a problem and then develop solutions at decreasingly low levels of detail until they are left with actionable sets of schematics and instructions. This, however, has always been the domain of human engineers as the engineering method is designed for — and assumes the presence of — human-level intelligence. Indeed, even recent advances in automation have only been able to recreate part of the conceptual design stage of the engineering process. The current state of the art involves knowledge-based engineering systems that can improve a human engineer’s efficiency and generative design systems that can completely solve low-level part design problems; both of which require close human supervision. Neither of these kinds of systems, however, have the high-level design reasoning or agency of a human engineer and so they are unable to approach unfamiliar or more complex design problems. Additionally, trusting their designs requires extra verification and testing on the human engineers’ part. The objective of this work was to take a step towards developing these high-level design capabilities by introducing methods and systems from the field of artificial general intelligence and implementing them in a small technology demonstrator. The resulting prototype, EDA-0, is capable of reasoning about basic conceptual design problems, identifying viable solutions, and outputting a human-readable trace of its decisions. It can also be further modified to include cooperation with a human engineer. Though limited in its current capabilities, this prototype lays important groundwork in problem decomposition, state representation, and human-computer cooperation. The hope is that the work performed here can form the foundation for a new generation of artificially intelligent engineers and designers.

Samþykkt: 
  • 30.1.2024
URI: 
  • http://hdl.handle.net/1946/46267


Skrár
Skráarnafn Stærð AðgangurLýsingSkráartegund 
A Foundation for Autonomous Conceptual Engineering Design.pdf1,62 MBOpinnHeildartextiPDFSkoða/Opna