• GET Service

Co-funded by the 7th Framework Program of the European Union
 

Progress beyond the state of the art

 

Transportation planning and control
Transportation planning and control is tackled in earlier European projects, including FINEST, EURIDICE, SMARTFREIGHT and SuperGreen. In addition the project partners of the GET Service project, in particular PTV has off-the-shelf solutions for (green) transportation planning.


These projects and research on (green) transportation planning algorithms in general, has advanced the state of the art to a stage at which it is possible to use historical data for more efficient planning (Van Woensel et al., 2008), to do green route planning by incorporating CO2 emission into the route planning criteria (Bektas and Laporte, 2011; Jabali, Van Woensel and de Kok, 2011) and to do multimodal planning of transportation routes (Jansen et al., 2004). In addition, the state of the art enables the use of real-time data to detect disruptions in the transportation plan due to, for example, traffic congestions that are currently present on the planned transportation route (Ichoua, Gendreau and Potvin, 2000; Jabali et al., 2009).

The GET Service project aims to advance on the state of the art in transportation planning, by aggregating real-time and planning data from multiple data sources. Thus enabling more precise and predictive planning, while also including transportation related tasks in the planning. For example, by combing data about the current situation on the road with the weather forecast, the presence of roadworks and the routes planned by other vehicles, a prediction can be made about the traffic situation on the road at the moment in the future at which a truck plans to pass that road. Further combining information about the average transfer time of goods from the truck to the ship and the administrative tasks surrounding that transfer, enables the truck driver to plan his route more precisely.

Service development and composition
To our knowledge, software service composition was not yet studied for the specific domain of transportation and logistics services. Prior research has focused on transportation and logistics planning, but not on automated control. Automated service composition has been studied in numerous publications in an abstract, domainindependent setting (Rao and Su, 2005; SUPER Consortium, 2009) . When studied in full generality, i.e., if we want to compose arbitrary software services, automated composition suffers from a high complexity, and no convincing practical solution has yet been created. On the other hand, if we do not want to compose arbitrary services, but only highly standardized building blocks (as in the LEGO modeling world), the problem becomes relatively simple. However, over-standardization limits the flexibility of a company in which of their existing services could be composed with the services of other participants in the GET service platform, which in turn can seriously impair the adoption of the platform in the market.


Therefore, one major research challenge is to strike the right balance between flexibility and standardization in order to achieve on-demand composability as well as a low entry barrier to the platform. This will be done by tailoring existing techniques to this application domain and standardize whenever necessary. Moreover, existing techniques for composition must be extended to take domain specific requirements into account, such as various forms of service constraint descriptions (e.g. legal constraints), as well as location-, time- and cost- awareness, all of which have not yet been sufficiently considered in relation with service composition.


Furthermore, to achieve the above mentioned flexibility, we will develop support functions for creating new transportation services to facilitate composability with existing services of other participants in the GET service framework. To this end, we will leverage prior work such as similarity search (Dijkman et al., 2011), process configuration (M. La Rosa et al., 2011) and extend and adapt them to the specific requirements of this application domain. We will also investigate whether techniques for generating adapters to make existing services composable (Gierds et al., 2010) can be made applicable to our use case. Few of the mentioned techniques in this paragraph have been rigorously tested in practical scenarios so far, none of them in the domain of transportation and logistics services.

Service composition orchestration and reconfiguration
A number of industry-strength service composition orchestration engines exist, also known as orchestration engines. Such engines are also used for transportation control in European Projects, such as the EURIDICE project. Currently, these engines support mechanisms that are a necessary precondition for enabling reconfiguration of service composition, including: cancellation of running tasks, roll-back of transactional tasks and compensation of already performed tasks. However, these mechanisms can only be used if the service composition explicitly models when they are invoked. In the context of this project, we aim to develop mechanisms that are automatically invoked in the correct way, depending on how a transportation plan and, therewith, the corresponding transportation control structure is changed.


Research has been performed in the area of dynamically changing service compositions. The outcomes of this research enable changes to a service composition. Changes can be made by replacing individual services with another service depending on quality of service characteristics (Ardagna and Pernici, 2007). Changes can also be made by removing or adding individual services (Casati et al., 2000) either from a running composition instance or a set of running instances. The most advanced systems enable the adaptation of a service composition, using pre-defined adaptation ‘patterns’ (Dadam and Reichert, 2009). Related to research efforts aiming at dynamic reconfiguration of service compositions are the efforts aimed at dynamic (on-demand) composition of service-oriented business processes. Several approaches from European Projects exist that differ in the complexity of composition that they support, ranging from bilateral dynamic service outsourcing (Grefen et al., 2000) to multilateral process networks (Grefen et al., 2009). These approaches have been prototyped in logistics, insurance and manufacturing application settings where demands for agility in business are increasing. Some approaches pay explicit attention to advanced transaction support (Grefen et al., 2000; Wang et al., 2008) comparable to that discussed above for orchestration engines.

The GET Service project aims to advance the state of the art in dynamic reconfiguration of a service composition, by exploring the reconfiguration patterns that should be supported by transportation and logistics service compositions and developing specific support for those patterns. Ideally, the project develops dynamic reconfiguration mechanisms that support arbitrary adaptations to a service composition.

Information aggregation
Information aggregation is typically done by ‘Complex Event Processing’ (CEP) engines (Luckham, 2011). These engine are able to efficiently aggregate high-volume high-detail events that are broadcasted on a technical level to events that are meaningful on a business level, based on predefined aggregation functions. The FINEST project aims to develop such an engine. As a first step towards doing that. It present a highly detailed state of the art analysis of CEP engines and related technology (Engel et al., 2011). CEP engines use definitions of aggregation functions to do their work, typically using a language that is specifically created for defining these functions.

A central problem in this context is that event data is often object-centric rather than process centric.Existing work on complex event processing hardly considers an explicit reference to individual service composition instances. The Process Mining Manifesto identifies the efficient and effective mapping of such event data to service composition instances as one of the core challenges in this area (cf. van der Aalst 2012). The GET Service project aims to advance the state of the art in information aggregation, by exploiting the planning and control structures that are used and constructed by the GET Service platform (e.g. Figure 4) to facilitate the targeted generation of business-level events. A detailed analysis of the possibilities that these structures provide will be performed during the project. However, two directions that will at least be explored are the following. First, the possibility to (automatically) generate aggregation functions, based on known technical-level data and desired business-level data, will be explored. Second, the possibility to derive missing business-level data from technical-level events will be explored. Some sensory events might be missing, due to communication failures or other exceptional reasons. Methods to deal with incomplete and exceptional events need to be analyzed and adapted to the transportation and logistics domain, a.o. based on behavioural constraints (Weidlich, Polyvyanyy et al. 2011, Weidlich, Ziekow et al. 2011).

Continue with the Expected Impact