A cerebellar operant conditioning-inspired constraint satisfaction approach for product design concept generation

Conceptual design is a pivotal stage of new product development. The function-behaviour-structure framework is adopted in this stage to help designers search design space and generate conceptual solutions iteratively. Computer-aided methods developed within this framework will yield significant insight into facilitating the cognitive activities of designers. In order to solve the mapping process from behaviours to structures which is a typical constraint satisfaction problem, a cerebellar operant conditioning-inspired constraint satisfaction approach is proposed in this paper. The design constraints-driven operant conditioning and its regulation mechanism by the cerebellum are analysed for the first time. Proposition logic is applied to transfer the constraint satisfaction problem into a propositional satisfiability problem while an undirected graph is utilised to model design space. Inspired by the modularised cerebellar structure, a modularised constraint satisfaction neural network is constructed to determine the satisfiability of design problems. Conceptual solutions can be generated by clustering the embedding of nodes in this network. The proposed approach imitates the design constraint-driven operant conditioning to narrow down design space without assigning specific values to design components. It reduces design iterations and avoids combinatorial explosions during conceptual design.

Automated summary

Conceptual design is a crucial stage in new product development, and the function-behaviour-structure framework is adopted to aid designers in searching and generating conceptual solutions. Computer-aided methods within this framework facilitate cognitive activities and propose a cerebellar operant conditioning-inspired approach to solve the mapping process from behaviours to structures. A modularised constraint satisfaction neural network is constructed, inspired by the cerebellar structure, to determine the satisfiability of design problems and generate conceptual solutions by clustering embedded nodes. This approach imitates design constraint-driven operant conditioning, reducing design iterations and avoiding combinatorial explosions in conceptual design.