Wednesday, June 1, 2011

Towards Grammars for Cradle-to-Cradle Design (Doug Fisher & Mary Lou Maher)

Cradle-to-cradle (C2C) design (McDonough & Braungart, 2002) recognizes that nothing short of full recycling of materials with no degradation in material quality is necessary for long-term planet sustainability. C2C advocates looking to the natural world as an ideal model of recycling, where organic materials are continually recycled through processes of decay and growth. They propose design methodology that separates biological cycles and synthetic-material cycles, enabling biological material to be reclaimed through natural processes without synthetic and toxic residue, and enabling the full reuse of synthetic material through technological processes so as to eliminate resource depletion and toxic poisoning. While domain specialists define and elaborate these interacting cycles of materialuse, there are ample opportunities for Artificial Intelligence techniques to manage the complexities and articulate C2C design processes.

C2C goals suggest the necessity of holistic approaches that design the cycling of material, and include attention to the energy required to maintain the cycles. Reuse of material from one product line can be cycled back to the same product line or another product line. In reality, this already happens through normal recycling (e.g., plastic bottles are recycled into park benches), but the sources and targets of recycling are typically identified after major design decisions, leading to inefficiencies, material loss, and degradation. By designing C2C product families, reuse cycles can be made more efficient, with known and predictable trajectories for reused material. The figure illustrates that while there may be materials that come from outside a family and there are materials that are byproducts of the family production, a family design would seek to minimize these and to exploit them in a still larger context. That is, product families are dense subgraphs within a larger network.

Since grammars have been used extensively to model design processes, we are considering grammars to model C2C design of products and product families; productions of a grammar can represent operators for composing objects, as well as operators for decomposing objects (e.g., for purposes of recycling). We are defining classes of grammars that are aligned with C2C design desiderata. For example, a class of reversible designs may be represented by augmented grammars that limit artifacts to those that can be decomposed into basic parts with energy that is no greater than the energy required to compose the artifact initially; an example of an artifact that is not reversible might be a leather shoe that is tanned with a toxic tanner – the energy required to apply the tanner may be trivial, but the energy required to separate the tanner from leather at the shoe’s end of life may be horrendous.

Grammars, heuristic search and various forms of machine learning are highlighted as critical in grappling with the complexities of C2C design under many tradeoffs and constraints involving material and energy.

A preprint of the paper can be found at:

The presentation can be found at