Thesis summary This report contains the detailed course of designing an ontology that formalises the domain knowledge of City Logistics and then facilitates relevant agent-based modelling. Validation, trials of application and recommendation on future work are also included. The formal output of this work is an ontology document edited with Web Ontology Language (OWL) and is named as City_logistics_ontology.owl. Motivation City Logistics is a discipline specialised to cope with the sustainability problems encountered in urban freight transport. A key characteristic of it is the heterogeneity of the stakeholders involved. Besides the traditional logistics actors like shippers, carriers and receivers that share consistent interests (i.e. price and quality), City Logistics highly respect the interests of administrators and citizens that care more about the social welfare. To reach an optimal balance between private and public benefit, it is necessary to understand and in turn forecast the behaviour pattern of different groups. In recent years, agent-based modelling has been practiced as an unconventional tool to fulfil this task for its strong capability on capturing the dynamic behaviour of individual stakeholders and their interconnections. Referring to other domains (e.g. energy system) where the application of agent-based modelling is relatively mature, a following urgency is to achieve interoperability and in turn reusability between models via introducing formal ontologies as shared templates with which developers can standardise their models. Driven by this motivation, this thesis work focuses on designing a competent ontology that formalises the domain knowledge of City Logistics. Research goal The goal of this research is to develop an ontology for the City Logistics domain and in turn: - Formally and systematically describe the domain; - Help to analyse domain knowledge; - Facilitate domain-specific (agent-based) modelling with reusing the knowledge provided by the ontology. Methodology The methodology used to steer the ontology-building is customised on the basis of the one proposed by Natalya and Deborah as shown in Figure 1 (Natalya, et al., 2001). At first, the domain and scope for ontology development is defined, latter terms that are of domain-specific interest need to be collected from relevant literatures. After a series of sorting and refining, these terms are used as bricks to build the ontology. Finally a hierarchical structure together with the properties of its multi-level subclasses will be generated on the basis of these terms. Relevant case studies are performed to check the comprehensiveness as well as representativeness of the ontology. Based on the result of validation, recommendation on future work will be given. 1
Thesis Summary: An Ontology for City Logistics Determine the domain and scope of the ontology Enumerate important terms in the ontology Define the classes and the class hierarchy Define the properties of classes Validation & Evaluation with case study Recommedation on future work Figure 1. Methodology for ontology development, adapted from (Natalya, et al., 2001) Main course of ontology-developing Following this approach, we at first define the scope of this ontology as deliveries from end depots (excluding UCCs) to urban premises/homes by freight vehicles, and deliveries from retail premises to home by freight vehicles, and their return trip along with auxiliary logistics activities that will influence City Logistics performance. A motivating scenario describing the consequence of underestimating the heterogeneous decision-making process of stakeholders involved in City Logistics projects is given to highlight the importance of having an ontology that can help model developers to capture these complex, and usually implicit relations in their models. Large effort has been made to enumerate the terms that are of interest, and will probably be mentioned in the ontology. Almost each term can find its root from literatures by influential scholars in the City Logistics domain. Some customisations are made to neutralise the contradictions between different authors and in turn reach consistency. The terms listed out are used as bricks to build the hierarchical structure of the ontology, following the middle-out approach which advocates determining the most salient terms at first and expanding the structure from them. Starting from Stakeholder for its core status in the domain of City Logistics, another six top classes as shown in Table 1 are developed in a logical order. Table 1. Top classes in the City Logistics ontology Class Objective KPI Activity Measure Resource 1 2 Explanation Domain-specific objectives stakeholders want to achieve, and can be generally classified into economic, environmental and social objectives. The indictors that show the extent to which the objectives are fulfilled, are classified in accordance with the objectives (i.e. economic, environmental, social). Logistics activities performed by private actors to fulfil trading, cover ordering, validating (planning), (on-board) transiting 1, off-board transiting 2, loading, unloading, storing, breaking (splitting), consolidating. Solutions that can be performed by stakeholders to achieve objectives, generally fall into governance measures implemented by public authorities and private (company-driven) measures. Entities that can be used to perform activities or implement measures, ranges from monetary ones including public and private fund to non-monetary ones broadly covering infrastructure, physical location of private actors, freight vehicle, load unit, handling equipment, goods, fuels, personnel and software, etc. Movement of freight vehicles between two physical locations of private actors. The transiting between parked vehicle and the end destination (e.g. a specific section in a department store). 2
R&D Research and development on the CL domain, is dedicated to improving measures and resources. The main function of this class is to act as a library index where domain-specific literatures and other efforts can be recorded and categorised. Figure 2. Primary relations among top classes in the CL ontology Together with Stakeholder, these seven top classes are highly interlaced as shown in Figure 2. These interconnections among classes/individuals are exactly the so-called object properties that specify the relations between objects. As depicted in Figure 2, stakeholders have objectives that can be achieved by implementing measures. The measures can influence the KPIs of these objectives via intervening in the corresponding logistics activities performed by specific stakeholders. For example, a measure of restriction on loading time implemented by administrators who want to alleviate local congestion will at first force carriers and receivers to reduce time for unloading on kerbsides, and in turn shorten the average occupation of parking/loading slots. Thanks to it, double-park caused by fully-occupied kerbsides can be avoided. The ultimate effect is an increase in average travelling speed (KPI of congestion reduction ) contributed by less blocked streets. As a side effect, measures can simultaneously impair certain stakeholders by hampering their objectives. Using the same example, we can perceive stakeholders such as carriers will probably be negatively influenced by the restriction because they are expected to receive more parking tickets or use more crews on-board. This will directly lead to a higher logistics cost, which is contradictory to their objective. Some measures can complement other measures and thus are recommended to be implemented together with their beneficiaries as a measure package. For example, when local authorities have implemented time-window, the higher level authorities need to harmonise those time-windows so that carriers can still efficiently arrange multi-drop roundtrips with destinations in different cities. Certain objectives can t be achieved directly and the fulfilments of them rely on other objective. For instance, competitive retail industry is an indirect objective that relies on congestion reduction, nuisance (i.e. noise, vibration, visual intrusion and safety threat) reduction and valuable area (i.e. historic centre and public space) protection. These objectives aim jointly on improving the attractiveness of cities, which is exactly the key to the prosperity of local retail industry. In order to practice measures or activities, stakeholder should be equipped with the necessary resources (the support relation from Resource to Activity and Measure in Figure 2). A typical example can be the carriers 3
who use their unique resources as vehicles and drivers to perform activity as transiting from one physical location to another, and to practice measures like improving driver performance. The allocation of resources also determines the demand-supply relations among private actors. For instance, a shipper contracts a 3PL carrier just for their trucks and drivers, while the 3PL carrier can acquire money the resource it wants from its customer. It s the same story for shippers who sell goods and receivers who order goods. Inspired by (Lian, et al., 2007), we define activities in the ontology as a continuous flow where the end of one activity will trigger the start of others. For example, when a transiting is finished, there could be several follow-ups to be triggered depending on different situation. If the transiting is an outbound trip for delivery, then an unloading will be triggered. If the trip is for picking returns from retailers, then an off-board transiting will be triggered instead. Conceivably, it is possible to roughly reproduce the typical process flow of urban freight transport with these activities. Besides object property connecting one object to other objects, each object is also specified with data properties which quantitatively describe its attributes. An intuitive example is the gross vehicle weight of a truck. Validation An ontology is a conceptual model of a specific domain. Thus, like any other models, validation is required to test and evaluate the performance of it. With the finish of attaching data-properties, the ontology is more or less there. Case studies are then required to check the representativeness as well as comprehensiveness of it. Following clearly defined criteria, one theoretical modelling case and two real-life City Logistics projects are cited to see the binary compatibility of the ontology (i.e. the compatibility with the real world and the compatibility with the agent-based model). After a detailed comparison between the content provided by these cases and the corresponding (if any) elements in the ontology, the result turns out to be encouraging. The ontology proves to be able to systematically cover a large share (91% of the topics in the cases) of the domain knowledge and its ontological structure mirrors the agent-based model well. The main drawback lies in the inherent inability to capture dynamic issue like routing algorithm. However the basic information required for routing planning has been covered (e.g. address, GPS coordinate of ship address and delivery address, vehicle capacity, etc). Application The potential application of the City Logistics ontology is introduced after the validation. A more static usage of the ontology is to act as a database for reference where the users can be informed about the knowledge structure of the domain as well as the relations among concepts. The ontology can also work dynamically as a program (like Excel) that is able to help sort and analyse the knowledge it contains by automated categorisation as well as query-answering. These applications tend to be more valuable when the size of the knowledge contained is very large and the interrelations among concepts are complex. Recommendation on future work The following work is recommended to be launched from three perspectives. At first, every class in the ontology can be enriched continuously due the extensiveness as well as ever-lasting expansion of the domain knowledge represented by the ontology. Sometimes it s even possible to directly import external ontologies to replace the less-specified classes in the City Logistics ontology. Some data properties such as engine of vehicles can be converted into classes that are able to get further classified as well as specified with relevant data properties. Moreover, this kind of conversion can even introduce new semantics. An intuitive example is that terms such as 4
reputation and address can be connected to a new activity as, for instance, Choosing_supplier via a new object property as, for instance again, depending_on after the conversion. This formulation will explicitly indicate that reputation and physical location (the distance between the bases of shippers to the receivers sites) are two important criteria during supplier-choosing. Secondly, certain ontological restrictions on classes and properties can be further refined to improve the accuracy of the ontology, and this implies higher hardware demand. Thirdly, this ontology is still in a generic level due to the extensiveness of the domain it tries to represent as well as the relatively lower importance of instance-creating at the current phase. As a result, many classes just end up with sub-classes rather than specific individuals that are addressed with concrete data. Fortunately, the ontology itself can act as the template for instance-building since all the attributes that an instance should possess have already been stipulated upon the class it belongs to. The work left is just collecting and assigning data to the corresponding slots. Reference Lian, Park and Kwon. 2007. Design of Logistics Ontology for Semantic Representation of Situation in Logistics. Second Workshop on Digital Media and its Application in Museum & Heritage. 2007. Natalya, Noy F and Deborah, McGuinness L. 2001. Ontology Development 101: A Guide to Creating Your First Ontology. 2001. 5