High-Level Modelling and Solving for Online, Real-Time, and Multiagent Combinatorial Optimisation

This thesis presents a novel framework for declarative modelling and continual solving of dynamic combinatorial problems (e.g., planning the routes of delivery vehicles). This framework is then enhanced with a garbage collection mechanism, ensuring that the evolving problem's internal representation remains small as time progresses. Experiments show that garbage collection vastly improves runtime. This thesis presents another novel framework allowing experimentation with fairness in similar problems, which we show can achieve remarkably fair and resource-efficient solutions. Finally, this thesis presents one novel constraint propagation algorithm for variance and one for the Gini coefficient, which we show improve runtime.