Route Modelling for Gritting Vehicles a GIS-Based Approach for the Municipality of Rotterdam
Total Page:16
File Type:pdf, Size:1020Kb
Route modelling for gritting vehicles A GIS-based approach for the Municipality of Rotterdam Rik de Kleijn Msc Thesis Date: July 2018 Supervisor TU Delft: Drs. Wilko Quak Responsible Professor TU Delft: Prof. Dr. Ir. Peter van Oosterom Supervisor Municipality of Rotterdam: Christian Veldhuis 2 Colophon GIMA MSc Thesis July 2018 This master thesis is commissioned by the University of Utrecht, University of Wageningen, University of Twente and the Delft University of Technology as part of the Geographical Information Management and Applications (GIMA) Master of Science. Supervisor TU Delft: Drs. Wilko Quak Responsible Professor TU Delft: Prof. Dr. Ir. Peter van Oosterom External Supervisor Municipality of Rotterdam: Christian Veldhuis Contact information Rik de Kleijn [email protected] UU Student number: 3474445 ITC Student number: s6025862 3 Summary The municipality of Rotterdam is responsible for keeping local roads free from snow and ice to keep vehicles and pedestrians safe while travelling. This responsibility is primarily the prevention from roads becoming slippery, whereby specialized vehicles are being used for spreading salt on roadways. The expenditures involved in winter road maintenances are notable and exceed the 3 million Euros for an average winter. To decrease costs the municipality of Rotterdam is looking for the possibility to optimize gritting routes by means of a vehicle routing model. A vehicle routing model determines a set of routes, each performed by a vehicle that starts and ends at its own depot, such that all road segments are serviced. It is possible that these vehicle routing models may outperform own manual drawn routes, which are currently used for routing gritting vehicles. The first stages of digitizing routes and recalculate routing is important in the increasing use of ICT in winter service management. The main research question of this thesis is: “To what extent can the adaptation of a routing model decrease the driving time of gritting vehicles in the municipality of Rotterdam?”. It must be noted that the main objective is not to endlessly optimize gritting routes, but rather to adopt a routing model within the municipal organisation of Rotterdam. Two contributes are made: (i) to formulate a process for acquiring, cleaning and manipulating network data, which are stored in a network dataset with network capabilities, and can be used as input for modelling gritting routes and (ii) to develop a process for generating, tweaking and adjusting a routing model, using a real network dataset based on driver distances. A review of routing characteristics in the organisation show that the winter service organisation take care of all gritting responsibilities. The gritting vehicles start from one of two depots in Rotterdam and have to, in poor weather conditions, service all roads subject to gritting in a two-hour time window one the vehicles leave the depot. The two-hour requirement is held even when weather conditions are better. Salt capacity on vehicles is however not seen a constraining factor. In the literature review these characteristics are also indicated to be relevant and additional characteristics are added regarding the road network. That is, that a graph representation of a transportation network with one- way and two-way streets to be serviced and not every road segment traversed may need to be serviced. For the routing problem it is chosen to formulate it as a vehicle routing problem with a time- window. The vehicle routing problem is widely used and integrated within most commercial GIS applications. A GIS routing model workflow is created with ESRI’s ArcGIS and used to recalculate two gritting routes of the Hoogvliet neighbourhood in Rotterdam. The GIS routing model consisted of thirteen phases. The routing model begins with a process for acquiring, cleaning and manipulation asset management network data, defined in data input and pre-processing stage of the GIS routing model workflow. At the core the routes are formulated using the vehicle routing solver of ArcGIS network analyst, which uses a meta-heuristic tabu search based algorithm. In the post-processing steps the route output is generated and the route reporting and route visualisations are made. The routing output is then generated and tweaked to represent arc traversals of winter service vehicles. In the routing model workflow, a possibility is added that some of the data input and parameters can be changed throughout the process. To test the model on the gritting routes of Hoogvliet, first the current routing is loaded into the model. The results compared to three scenarios. The first scenario is the new routing scenario. This scenario does not impose any restrictions on any of the two vehicles or the sequence of routing. The algorithm is set to find the most efficient routing based on total vehicle travel time. The result found is that the total travel time increased by 5.4% in comparison to the current routing. The sectoring scenario creates a categorisation of service roads based on administrative districting to create more compact routing. 4 The model run had an increase of total vehicle travel time of 8.4%. The last scenario is the node reduction scenario, in which nodes are reduced between two gritting route junctions to reduce complexity. The node reduction scenario did outperform the current gritting routes by 5.3% but did leave some arcs ungritted and unconnected inside the routing. The results shed light on the use of current GIS infrastructure to help in adopting a routing model for gritting vehicles. Although the routing model can be used to decrease driving time, it was tested at only two gritting roads in Hoogvliet. For implementation the routing calculations have to scale to 30 main gritting routes. To implement some issues remain on both technical and organisational issues. The algorithm is not specially designed for large arc routing problems and has poor performance both in speed & accuracy. Districting is necessary to account for these issues. Furthermore, the network dataset cannot be simplified to reduce model solving complexity. From an organisation perspective, routing is labour intensive and quite specific. If one wants to proceed in recalculate the routing just once, one may want to consider these points before adopting a GIS routing model within the organisation. This thesis therefore makes four recommendations for adapting gritting routing within the municipality of Rotterdam. The first recommendation is to obtain or outsource the GIS software which is used to solve the formulated gritting routing problem. The current software which is available is not sufficient for solving large gritting routing problems. Secondly, it insists on the important notion that the routing must be closely created with the help of the winter service organisation. Route creation is an iterative process which require devoted attention by users. Thirdly, create a storage model for winter service asset management data and lastly make sure that the route creation is embedded in the organisation to ensure a sustainable routing solution. 5 Acknowledgements This thesis is the final project of my master Geographical Information Management & Applications (GIMA). For my final thesis I wanted to tackle a problem which touched both the managerial side of geoinformation management, as well as the more technical side by using GIS applications. With the help of my internship position at the municipality of Rotterdam I found a topic in vehicle route modelling which satisfied both criteria. Being able to write my master thesis within the geo advisory team of the municipality greatly helped in finding both theoretical and practical gratification in completing my master thesis. At first, I would like to thank my colleagues from the municipality of Rotterdam for their enthusiasm. A special thanks to Rob Poll- van Dasselaar, who introduced me within the municipality and Christian Veldhuis, who was a great support to me with his practical view on the organisational side of the project. Also, I would like to thank Willem Bouwmeester for our meetings regarding the WINTER project and his invitation to join the yearly test round for the staff routing for the upcoming winter. Of course, thanks to my friends and family which were of great support. A special notification to my dad for his support and to Sietske Tjalma for proofreading my thesis. I would also like to express my gratitude to my supervisor Wilko Quak from the TU Delft for his supervision and support. His enthusiasm and our meetings were of great value moving my thesis forward. Lastly, I would like to thank Peter van Oosterom for his final remarks. Rik de Kleijn Utrecht, June 2018 6 Table of content Summary ..................................................................................................................................... 4 Acknowledgements ...................................................................................................................... 6 List of figures ................................................................................................................................ 9 List of tables ............................................................................................................................... 10 1. Introduction ........................................................................................................................... 11 1.1 Introduction ................................................................................................................................