Artificial Intelligence for Automated Design of Elevator Systems

Configuration and design of complex products represents a challenge in many application fields. The designer must take into account many different aspects and make decisions typically driven by experience while taking into account performance constraints and costs. Methods and tools for design automation represents a viable solution to such complex decision problems, giving also the possibility to optimize the performance of the final product on particular context-driven aspects. Artificial intelligence (AI) algorithms can help in dealing with complexity and enhance the current tools by supplying solutions in feasible time. My research is concerned with the development and testing of different artificial intelligence (AI) techniques to automate the design of elevators. Elevator design is a problem with many interesting aspects like the need to deal with a hybrid search state space (continuous and discrete variables) constrained by design requirements and safety regulations. The study, design and integration of AI techniques in this particular application field can provide the end user with design automation tools that output feasible solutions within acceptable computation times. My research considered AI techniques such as special-purpose heuristic search, genetic algorithms and constraint satisfaction to solve elevator configuration problems. I tested them considering different setups and parts of the whole design process. I have also implemented a tool L IFT C REATE , available as a web application. L IFT C REATE leverages the findings of my research to automate the design of elevators and, to the best of my knowledge, there is currently no similar tool publicly available from either academia or industry that provides the same level of design automation.