In this talk, we present our work on two real-world logistics problems.
The first problem was brought to our attention by our industry partner PTV Group. It is an extension of the classical territory design problem and arises when companies operate a field service workforce to provide recurring services at their customers’ sites. Besides partitioning a set of customers into service territories, as in the classical problem, customer visits must be scheduled within the planning horizon subject to customer-specific visiting requirements (e.g., visiting frequencies). We refer to this problem as the multi-period service territory design problem and lay the focus of this talk on its scheduling component. We introduce the relevant planning criteria and present a mixed integer programming (MIP) formulation. Moreover, we propose two solution methods, namely a location-allocation heuristic and an exact branch-and-price algorithm. Experiments on real-world data sets prove the effectiveness of both methods: The location-allocation heuristic clearly beats the previous solution method of PTV Group in terms of solution quality. The branch-and-price algorithm yields significantly shorter running times compared to solving a compact MIP formulation with the state-of-the-art MIP solver Gurobi.
The second problem is tackled as part of our cooperation with Robert Bosch GmbH. It arises at large logistics centers and production sites, where a high number of trucks arrives each day to deliver or pickup goods. The problem consists of dynamically assigning the trucks to compatible docks taking into account continuously incoming events about expected and actual truck arrivals as well as the current state of the docks. It can be modelled as a flexible job shop scheduling problem extended by additional aspects, such as the possibility to reject trucks at the gate and the capacity of a parking lot. We refer to the extended problem as the dynamic gate and dock management problem. Since the problem must be solved many times throughout a day within an interactive decision support system, the running time of an optimization algorithm must be within few seconds. To meet this requirement, we develop and compare two different solution approaches: A mathematical programming based approach and a constraint programming approach. We evaluate both approaches on a large set of realistic test instances, which reflect the dynamic behavior throughout a day, and we discuss their strengths and weaknesses.
The work on multi-period service territory design is joint work with Jörg Kalcsics, Anne Meyer (FZI & TU Dortmund), Stefan Nickel (FZI & Karlsruhe Institute of Technology), and Martin Pouls (FZI). The work on dynamic gate and dock management is joint work with Anne Meyer (FZI & TU Dortmund).
Current 2019 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996
