Formulating combinatorial optimisation problems for quantum hardware usually requires encoding conditions as QUBO penalty terms, a task demanding expertise most domain practitioners lack. Despite this difficulty, no tool-assisted method exists to translate a domain model into a QUBO formulation, so practitioners must still hand-craft their hamiltonians. We address this gap with USE2QUBO, a plugin for USE OCL, a Unified Modelling Language (UML)-based specification environment that automates Quadratic Unconstrained Binary Optimisation (QUBO) hamiltonians and Q matrix derivation directly from a domain model, applying AutoQUBO sampling against USE’s own OCL evaluator. Object Constraint Language (OCL) invariants, written at the level of domain entities, map to QUBO penalty terms via a systematic sampling method heavily inspired by the work in AutoQUBO [13], without requiring expertise at the modelling stage. We applied the plugin to two structurally different proof-of-concept problems: a waste collection routing problem and a maximum clique selection problem to demonstrate how the plugin generated hamiltonians for each problem. All OCL invariants from each example scenario are validated with the USE tool [11]; both problems were able to map to QUBO hamiltonians, and both Q matrices were derived by the plugin with no coding required. Our approach raises the abstraction level above code-level tools and allows domain experts with some software engineering background to derive QUBO expressions from their domain-specific problems by leveraging software modelling and OCL.

We want to thank the authors of [13, 14], since their work was crucial in enabling the plugin's development and approach.

USE,OCL,QUBO,quantum modelling,model-driven engineering,combinatorial optimisation,VRP,waste collection routing,maximum clique,smart cities

• Computer and Information Sciences - Natural Sciences
• Electrical Engineering, Electronic Engineering, Information Engineering - Engineering and Technology