Analysis, modeling and solution of the concrete delivery problem
Introduction
Scheduling concrete mixer vehicles to deliver concrete from concrete-producing depots to concrete-demanding customers introduces some new combinatorial challenges from the viewpoint of traditional vehicle routing and scheduling problems. Besides the sheer size of usual practical problem instances of the concrete-producing/delivering industry we have to take into account constraints and properties that are not considered in standard vehicle routing and scheduling problems.
Most of the additional constraints necessary come directly from the material to deliver. Ready-mixed concrete is a perishable product, which has several ramifications for our problem. First, concrete may only reside in the concrete mixer vehicle for a certain amount of time before it looses quality for the customer and eventually hardens and can even destroy the barrel resulting in disproportional maintenance costs. Second, it is not advisable that vehicles transport a non-full load of concrete, because this could result in an increased rate of hardening of the concrete. This property prevents delivering concrete to more than one construction site before refilling the vehicle at a depot (delivering more concrete to a construction site than the demanded amount poses usually no problem or costs). Finally, if more than one delivery of concrete is necessary to satisfy the demand of the corresponding customer then the temporal spacings between the consecutive deliveries may not exceed certain limits (time lags) as the concrete could partially harden at the construction site before the subsequent supplies arrive.
Besides above mentioned complications for the scheduling part directly induced by the perishability of concrete, typical concrete delivering scenarios differ further from standard vehicle routing problems in the following aspects. Usually, we cannot expect total homogeneity of the vehicles. In our considered practical applications the available vehicle fleet contains concrete mixer vehicles capable of only transporting 2 m3 of concrete to vehicles that can deliver 8 m3. The demands of construction sites often exceed the vehicle capacities and thereby require chains of subsequent deliveries. Furthermore, there exists usually more than one depot and there is enough concrete demand during a given working day calling for decisions from which depot to deliver to which customer.
Altogether we have a highly complex scheduling problem at the intersection of several research areas, such as logistics, supply chains and just-in-time production. We will try to investigate the concrete delivery problem in a general form, but still being able to solve problem instances coming from practice. For this aim this paper is organized as follows. In Section 2, we will give a short review on existing papers on the subject. From Section 3 on we will formulate a general model of the concrete delivery problem by first defining a notation for the input parameters of the problem. Section 4 continues with the definition of a network flow model based on the input data and results eventually in a mixed integer programming model (MIP) which describes formally the general concrete delivery problem that we consider and could be used to solve such problem instances. In Section 5 we present a certain general local search approach, which we utilize to solve our problem in Section 6. The paper finishes in Sections 7 Computational results, 8 Conclusions where the computational results are given and conclusions drawn.
Section snippets
Related work
The logistical problems arising in concrete delivery seem to represent a rather young research field. Following are most publications in this area that are known to us.
Durbin researches in his dissertation [6] the concrete delivery problem both practically and theoretically. The result is a decision support system which is used in practice and incorporates different models described in his thesis. The problem he considers is a special case of ours as it assumes homogenous vehicles. He mostly
Problem parameters and notation
Since the concrete delivery problem that we consider is similar to the vehicle routing problem with time windows (VRPTW), we have based our notation especially on [5], which contains the prevalent notation for VRPTW problems (for an overview over VRPTW problems and the plethora of algorithms we refer to [3], [4] besides the aforementioned paper). Thus, we will simply call the concrete mixers just vehicles and the construction sites customers.
We consider a static model where all data are
Flow network
Given the input parameters defined in the last section we can now define a graph or flow network on which the mixed integer programming model will be based on. The basic idea of the graph is that we represent each possible supply/delivery of a customer, each possible reload of a vehicle at a depot and each starting and ending point of a vehicle as a node and then add all edges to the graph that correspond to a part of a route that could be contained in an optimal solution of our problem. We
Generic local search approach
Because of its size for realistic instances from practice the aforementioned MIP model does not permit generation of an optimal solution using CPLEX. Therefore, we have tried to combine this model with a certain local search approach motivated by Kilby et al. and Schrimpf et al. [9], [12].
The basic idea of our local search approach is the following. Consider a generic local search algorithm . This algorithm will start with an initial solution s – the incumbent, the current best solution – and
Heuristic for solving the subproblems
The heuristic algorithm begins with fixing all depot node times to constants as follows: for all . This may already cut optimal solutions from the solution space, but makes subsequent algorithms much simpler and its impact on the solution quality seems to be rather small on bigger instances, where the depots can reload vehicles at a much faster rate than in our example in Fig. 1.
The core of the heuristic is presented
Computational results
To evaluate our algorithms we tested them on some real world data coming from a typical working week from a big concrete distributing firm which operates in several geographical distinct regions with different vehicle fleets. The real world data and constraints are modelled by our aforementioned MIP model and can therefore be solved with the preceding local search algorithm. All solutions with given run times were achieved on a 1700 MHz Intel Pentium M processor notebook with 512 MB RAM on the
Conclusions
The topic of this paper is to model and solve the logistical problem of delivering concrete from concrete-producing depots to concrete-demanding customers.
To the best of our knowledge our model and the problem itself is the most general one considered in the literature so far and describes the practical problem in our opinion intuitively and rather easily in a formal way. Unfortunately, the size of the model does not allow solving real world instances with an MIP solver. Therefore, we have
References (13)
- et al.
Resource-constrained project scheduling: Notation, classification, models and methods
European Journal of Operational Research
(1999) - et al.
Optimizing the schedule of dispatching RMC trucks through genetic algorithms
Automation in Construction
(2004) Towards a decision support system for the ready concrete distribution system: A case of a greek company
European Journal of Operational Research
(2004)- et al.
Record breaking optimization results using the ruin and recreate principle
Journal of Computational Physics
(2000) - H.-J. Böckenhauer, On the Inapproximability of the Metric Traveling Salesman Problem, PhD Thesis, RWTH Aachen, Germany,...
- et al.
Vehicle routing problem with time windows, part I: Route construction and local search algorithms
Transportation Science
(2005)
Cited by (63)
Resources synchronization in a full truckload pickup and delivery problem: An exact approach
2023, Computers and Operations ResearchCitation Excerpt :An Adaptative Large Neighborhood Search (ALNS) metaheuristic approach is proposed and efficiently tested on real instances. Asbach et al. (2009) present a mixed integer linear programming (MILP) model that combines allocation, routing, and synchronization decisions, considering technical conditions of ready-mixed concrete delivery. The proposed MILP is not capable of solving real instances.
Systematic search and mapping review of the concrete delivery problem (CDP): Formulations, objectives, and data
2023, Automation in ConstructionCitation Excerpt :17 papers formulated or adopted a formulation based on mathematical programming. The most used one is the one by Asbach et al. [16]. Considering that this formulation is also the most comprehensive one, as shown in subsection 2.5, we expected that the researchers will prefer it.
Factors affecting the price of recycled concrete: A critical review
2022, Journal of Building EngineeringModel formulation and calibration procedure for integrated multi-modal activity routing and network assignment models
2020, Transportation Research Part C: Emerging TechnologiesCitation Excerpt :In the SDVRP literature, to formulate the multi-visit feature, scholars have enumerated either the deliveries to a specific node or the links connected to the nodes. Enumerating deliveries is commonly applied in scheduling problems (see e.g. Asbach et al., 2009; Berghman et al., 2014; Maghrebi et al., 2014; Narayanan et al., 2015). Although both options allow revisiting nodes more than once, enumerating the nodes and link may be costly in the context of supernetworks and ARP-based models where they can significantly increase the solution space.
Heuristic algorithm for ready-mixed concrete plant scheduling with multiple mixers
2017, Automation in ConstructionAn adaptive large neighborhood search for the full truckload pickup and delivery problem with resource synchronization
2017, Computers and Operations Research