Conference Agenda

Waste collection systems are an important component of solid waste management systems. Scientific studies have discussed different advantages and disadvantages when employing a single waste collection system. In this talk, we introduce the Combined Waste Collection Problem which combines the two most common waste collection systems: Door-To-Door Collection System, where waste is picked up from households using trucks or other vehicles and Automated Waste Collection Systems, which transport waste from inlets through a network of pipes to a central collection point.

Combined Waste Collection Problem is a two-stage decision problem: First, decide for each customer if she/he is served either with a truck or with the pneumatic system such that the capacity of the pneumatic system is respected. Second, solve a capacitated vehicle routing problem for the truck customers and construct a minimum spanning tree for the tree customers. We present a holistic solution approach based on a set-partitioning formulation utilizing route and tree variables. Since there are exponential many variables, we solve the problem with a column-generation approach. The arising subproblems are an elementary shortest path problem with capacity constraint and a variant of the price-collecting Steiner tree problem.

The substantial traffic within urban areas poses challenges for both Logistics Service Providers (LSPs) and urban communities. One potential solution to alleviate traffic congestion and maintain cost-effective transport is the implementation of Two-tier City Logistics Systems (2T-CLS).

Given the extensive infrastructure requirements and the presence of multiple LSPs operating within a city, cooperation among these LSPs offers opportunities for both economic and environmental cost savings. However, establishing effective long-term cooperation among LSPs necessitates ensuring that each LSP has incentives to cooperate with others and feels fairly treated. This requires that fairness with regard to the distribution of the workload and the cost is already taken into account in the decision support systems.

Therefore, we introduce a service network design formulation for the planning of cooperative 2T-CLS over a horizon of several days. This formulation includes both workload and cost balance constraints, which can optionally apply either on each individual day or combined over the entire planning period. Moreover, we consider constraints on the limited daily flexibility in selecting services to achieve a certain regularity in the daily schedules.

In a numerical study, we demonstrate that cooperation leads to benefits in the form of cost savings and lower negative environmental impact. In addition, we show that too strict fairness constraints harm the entire coalition and that lower cost increases occur when fairness constraints are enforced over a multi-day period rather than on each individual day.

Crowdsourced delivery utilizes the services of independent actors. As opposed to traditional modes of delivery, availability and acceptance decisions of crowdshippers are uncertain and cannot be fully controlled by an operator. We consider a setting in which an operator groups tasks into bundles and in which the resulting bundles are offered to crowdshippers in exchange for some compensation. Uncertainty in the crowdshippers' behavior who may accept or reject offers is considered via (individual) acceptance probabilities.

We introduce and study a model considering individual compensations that influence the acceptance behavior of occasional drivers. We model the latter via probability functions that estimate the likelihood of an occasional driver accepting a task based on available (historical) data, attributes of occasional drivers and tasks, as well as the offered compensation. We propose a mixed-integer non-linear programming (MINLP) formulation that simultaneously decides how to group tasks into bundles, which bundles are offered to which crowdshipper, and the compensation offered for each bundle. Our work is the first study to integrate compensation-dependent acceptance decisions in an exact solution method while also considering the option of offering bundles of tasks to individual drivers, minimizing the combined total expected cost of delivery and compensation.

We show that this MINLP can be reformulated as a mixed-integer linear program with an exponential number of variables. We present a column generation algorithm for solving instances of the latter. Our experiments show that the algorithm is capable of solving large instances.

With increasing e-commerce transactions and traffic congestion, handling large volumes of parcel deliveries efficiently in city logistics becomes more challenging. To this end, we aim to design a two-echelon distribution system where trucks, originating from a central depot outside the city center, deliver parcels to micro-depots in the city center. Smaller vehicles like cargo bikes handle customer deliveries from these micro-depots. We decide on the location of micro-depots, first-level routes of trucks from the central depot to the micro-depots, and second-level routes of cargo bikes from the micro-depots to customers to minimize the total costs incurred. Our model can also be adapted to allow direct shipment from the central depot to the customers.

From the perspective of a logistics service provider, it is essential to find a good distribution plan quickly for thousands of deliveries. To solve such large-scale problems efficiently, we propose a metaheuristic that integrates a set cover problem with an adaptive large neighborhood search (ALNS) algorithm. Our ALNS approach generates a set of promising routes and micro-depot locations in destroy and repair combined with a local search. We then utilize the set cover problem to find better network configurations during our search. We also develop a decomposition-based cluster-first, route-second approach to solve large-scale instances efficiently. We compare the decomposition approach with our integrated ALNS in terms of solution quality and runtime. We show the efficacy of our algorithm on well-known benchmark datasets and provide managerial insights based on a case study for the city of Munich.