Repeated Multi-Criteria Decision Making in Systems with Coupled Discrete Actuators

Multi-criteria decision-making methods are designed to rank a fixed set of alternatives once. Many technical systems, however, must repeatedly decide how to operate actuators that share physical states and a limited resource. In such systems, action scores become coefficients of a constrained selection. This article states four requirements for these scores, an interval scale, a zero at the predicted state of inaction, independence from the candidate set and indifference below a discrimination threshold, and shows that canonical scores violate the first three. Local modifications bring TOPSIS and PROMETHEE methods into compliance by anchoring the evaluation to the installation instead of the candidate set, and a 0–1 mixed-integer program selects the actions jointly. Software-in-the-loop simulations of a room with five actuating groups show that removing any of the first three requirements degrades closed-loop behaviour. Unlike rule-based baselines, the joint selection keeps the controlled variables in their target intervals when one actuator must serve two functions at once, and with the modified TOPSIS also under a binding resource limit. The constrained selection is found to determine whether to act, and then scores which action to take and at what cost. The two methods also differ in how closely their scores follow the declared weights, which shifts the balance between comfort and energy.