Conference Agenda

Transformation pathways provide information about which action to take in which building and

which year to transform residential building stocks. We define the problem of determining optimal

transformation pathways of residential building stocks and introduce a MILP model for this problem.

To accelerate the solving process, we propose different Benders decompositions of the original model.

For further acceleration, we introduce constraint modifications to generate Benders cuts faster as

well as valid inequalities for our Benders relaxed master problems (RMPs) and a tailored Benders

optimality cut strengthening technique. We further implement a primal construction heuristic for the

Benders RMP and an in-out method to improve the dual bound at the root node. The Benders cut

separation process is embedded into the branch-and-cut algorithm of the MILP solver Gurobi. Based

on the German Census 2011, we introduce instance sets to test the algorithmic implementations on

which we heuristically refine solution strategies based on aforementioned features. For instances with

50 buildings and a 20 years time horizon, the best-performing strategy yields relative MILP gaps of

5% while Gurobi attains relative MILP gaps of 20% and 10% if performing a warm start with our

construction heuristic.

1

Home services play an important role for business operations across various fields, such as after-sale services or craft businesses. In each case, technicians are sent to customer homes to complete on-site repair tasks. In this paper, we assume that workforces are heterogeneously qualified and that tasks require different skill levels. In contrast to other works in this field, we explicitly account for possible service failures due to a lack of technician skill. Tasks of higher complexity may remain unresolved, requiring revisits and rework. This consequently incurs additional resource investment and service delays. As we show in this paper, combining multiple dimensions (e.g., service deadlines, skill levels) during decision-making mitigates future service delays and resource investments. To this end, we introduce a cost-function approximation architecture to anticipate long-term implications at early decision points. In line with our objective to minimize overall customer delay, our method adapts its focus across multiple streams. This involves prioritizing the service of customers with imminent deadlines to avoid short-term delay, improving routing efficiency for increased customer visits, and assigning tasks to prevent service failures and rework. For various indicators (e.g., service delay, resource investments, rework), the results of a numerical study demonstrate that our anticipatory solution policy outperforms alternative benchmark policies that emphasize isolated goal dimensions. Our proposed policy achieves a fairer regional service distribution compared to benchmark methods while sustaining customer retention and ultimately enhancing operational performance.

Motivated by a real-world application in industrial storage yards, we consider the problem of scheduling predetermined container relocation moves by an automated twin-crane system. We show that this problem, which we call Twin-Crane Scheduling, is NP-hard (already in a simpler setting where some constraints are relaxed or dropped entirely). Further, we prove that even a polynomial-time approximation scheme cannot exist for this problem unless P = NP. However, we show that the optimal solution can easily be approximated within a factor of 2 through a simple list scheduling algorithm. Finally, we establish a connection between a generalized

crane scheduling problem and scheduling unit-time jobs on three or more machines under precedence constraints, the complexity of which has been a long-standing open problem.