Routing and Loading Bays Optimisation in Urban Areas Using Fuzzy Logic and Manhattan Metric

V disertaciji rešujemo problem statistično optimalnega števila in lokacij dostavnih mest ter optimalnega vodenja tovornih vozil v procesu mestne dostave. Za iskanje optimalnega števila in pozicij dostavnih mest v doktorski nalogi razvijemo večparametrični model idealiziranega mesta. Model obravnava urbano območje velikosti 1 km2 in temelji na metodi vzorčenja latinske hiperkocke za razporeditev strank vzdolž izbrane ortogonalne prometne mreže. Določanje statistično najprimernejših lokacij dostavnih mest temelji na algoritmu mehkega grozdenja c-sredin. Parametrične analize so pokazale statistično značilne geometrijske vzorce optimalnih dostavnih mest. Izkazalo se je, da so identificirani vzorci robustni glede gostote strank in gostote prometne mreže. Problem določanja dostavnih mest je intimno povezan s problemom pokritja. Obravnavan je pomemben mejni primer pokritja mreže z dokazom optimalnosti pokrivnega števila. V disertaciji je podana tudi optimalna konstrukcija limitnega primera na poljubni pravokotni mreži. Model določanja števila in lokacij dostavnih mest je nadgrajen s funkcijo optimalnega vodenja tovornih vozil. Ključna predpostavka modela so štiri vstopno/izstopna vrata v mesto ter pri vsakem obisku mesta največ trojna dostava iz dostavnih mest (t.j. trem različnim strankam). Glede izbire dostavnega mesta sta bila uporabljena dva principa, ki temeljita na rezultatu algoritma mehkega grozdenja c-sredin: (i) trdi – vedno se izbere dostavno mesto, ki je najbližje stranki, (ii) mehki –omogoča tudi izbiro dostavnih mest z nižjimi pripadnostnimi vrednostmi. Problem določitve optimalne transportne poti je rešen s pomočjo nabora izvirnih celoštevilskih linearnih programov. Rezultati simulacij za trojno dostavo kažejo, da skupne razdalje prevoženih poti niso značilno odvisne od števila potencialnih dostavnih mest, jih je pa mogoče pomembno zmanjšati z mehkih pristopom, vendar na račun daljših prehojenih razdalj. Izkaže se, da je najbolje uporabiti mehki pristop pri relativno velikem številu potencialnih dostavnih mest. Model je bil prilagojen tudi za reševanje konkretnega praktičnega problema na primeru mesta Lucca, v Italiji. Za obravnavano mesto smo izdelali usmerjen graf prometnega omrežja, potencialne lokacije dostavnih mest so bile določene na osnovi dostopnosti in prostorskih možnosti. Za generiranje povpraševanja po dostavah je bil razvit simulator, uvedena je bila tudi procedura rezerviranja dostavnih mest in simulacija čakalne vrste vozil zunaj mesta v primeru omejene kapacitete in zasedenosti dostavnih mest. Izračunani so bili časovni in energijski prihranki. S ciljem realizacije predlaganih rešitev je v nalogi predstavljen tudi koncept informacijskega sistema, ki temelji na uporabi mobilnih naprav in omogoča vodenje dostavnega vozila in s tem upravljanje procesa dostav v realnem času. Na koncu doktorske naloge so podane ključne ugotovitve in predlagani koraki za nadaljnje raziskovanje.The problem of statistically optimal number and location of loading bays and optimal routing of freight trasport vehicle into and outside the urban area is considered in the dissertation. To find the optimum number and position of loading bays, a multiparametric model of idealized urban areas was developed. Urban area of 1 km2 is considerd. The model is based on the Latin Hypercube Sampling method for simulating random distribution of customers along the othogonal transport network. Statistically optimal locations of loading bays are determined in terms of the Fuzzy c-means algorithm. Parametric analysis reveals significant geometric patterns for statistically optimal locations of loading bays. It turned out that the identified patterns are robust in sense of customers\u27 density and the transport network density. The problem of locating the loading bays on the network is highly related with the covering problem. An important limit case of the covering problem is discused in details and the optimal covering number is found and an optimal construction for any rectangle lattice is given in the dissertation. An upgrade of the model for determining the locations of loading bays was performed with the function of optimal vehicle routing. The key assumption of the model is the possibility to access the city only from four particular locatins, and that a particular delivery is made to at most three different clients per a visit of the city. Two different regimes, based on the results of the Fuzzy c-means algorithm, for selecting a loading bay were developed: (i) hard option - always use the closest to the customer lodaing bay, (ii) soft or fuzzy option – also the choice of loading bays with lower membership values is considered. The problem of determining the optimal transport route is solved via the set of original integer linear programs. The results of the simulations show that the total distance traveled does not depend on the number of loading bays. The total driven distance can be significantly reduced in soft option, but at the expense of longer walking distances. It turns out that the best option is to use the soft option and a relatively large number of loading bays. The model was adapted to solve the practical problem. The case study was made for the city of Lucca, Italy, which was covered by a directed graph. All potentially feasible, in sense of accessibility and spatial possibilities, loading bays were determined and a simulator for generating the demand for deliveries was developed, including the reservation procedure and simulating the waiting of vehicles (queue), outside the city in case of limited capacity and occupancy of loading bays. Time, distance energy and environmental savings were estimated. Finaly, with the aim to realize the proposed solutions, the concept of an information system based on the use of mobile devices is presented. The proposed information system enables to navigate the vehicles and to manage the delivery process in real time. At the end of the thesis key findings are summarized and future research challenges are proposed