Conclusions and Future Work
5.2 Future Work
The motivation and objective for the work presented in this dissertation has, as stated in Sect. 1.2, been to provide better facilities for conducting practical use of CP-nets. From the contributions described in Sect. 5.1 we claim that this has been achieved. Facilities have been developed and practical projects have shown the usability of these facilities. However, the work presented here can be improved in many ways. In this section we discuss a few ideas and directions for future work.
5.2.1 Web-Based Interfaces and Influence Nets
Currently work is conducted at George Mason University to change the in-fluence net tool to make it possible to specify timing parameters within the influence net tool instead of specifying these parameters via the web browser.
By using the influence net CPN simulator as an integrated component in the influence net tool, both input and output can be managed directly from the GUI of the influence net tool. In other words, the influence net tool itself will completely hide the fact that a CPN simulator is used internally for simulating influence nets with timing information. For some users this approach may be more appealing than having to use two different tools for creating and analysing influence nets.
Due to the interest from the US military in using CPN simulators for in-fluence nets, it may also be worth while to investigate how inin-fluence net CPN simulators can be integrated with other simulation models. The main reason why this may be relevant is that the US military put great effort into devel-oping a high-level architecture (HLA) [93]. HLA supports the combination of several simulation models into so-called federations of simulation models. In other words, the architecture makes it easy to create a distributed environment for running several interacting simulation models.
5.2.2 Monitors
In the paper on monitors we claim that the monitoring framework can probably be used for any discrete-event system formalism. This claim has to be investi-gated further in the future by investigating and evaluating other discrete-event system tools.
5.2. Future Work 37 As mentioned previously, the monitoring framework has been developed as a generalisation of the data collection facilities in theDesign/CPN Perfor-mance Tool. The Design/CPN Performance Tool has been used in several projects, and was integrated in version 4.0 of Design/CPN released in September, 1999. In contrast, only a prototype of the monitoring frame-work has been implemented and applied to a few CP-nets. Future frame-work should focus on completing the implementation of the monitoring framework. This in-cludes extending the GUI ofCPN Tools[23] which is the tool going to replace Design/CPN in the near future. In this tool the design of the GUI for the monitoring framework should use the new advanced user interaction techniques as described in [6, 5].
Future work could also continue our undocumented work onbatch monitors.
The monitors described in the monitoring framework are primarily intended to be applied during a single simulation. When using a model for e.g. performance analysis it is often necessary to run several simulations – possibly with slightly different parameters. To support monitoring during such batches of simula-tions we have considered how batch monitors could be designed. We believe that there is a great potential for batch monitors, and it should definitely be investigated further in the future.
5.2.3 Annotations
There are many issues regarding annotations that can be addressed in the fu-ture. First of all, the ideas on annotations presented here have not yet been used in practice due to lack of tool-support. It is of high priority to develop a prototype of a tool supporting annotations. Only by using annotations in practice on concrete CP-nets, it is possible to determine if annotations will make it easier for the modeller to maintain auxiliary information within a CP-net. The prototype of a GUI supporting annotations should be implemented inCPN Toolswhere the design of the GUI could use the new advanced user interaction techniques ofCPN Tools.
Further research is required on annotating arc expressions on arcs that eval-uate to multi-sets instead of single colours. The current proposal where all colours in the multi-set will get the same annotation may be too restrictive in practice. However, if the notation of annotations is too complex, then users may be reluctant with learning how to use annotations. Therefore, the limita-tions have been made to make the annotation notation as simple as possible – and thereby to make it easier to design tools with user-friendly support for annotations.
Guards are not considered in the proposal for annotations. The main reason is that guards are often added to restrict the enabling of transitions. However, guards can also be used for binding free variables. A future proposal for annota-tions should consider the ability of binding free auxiliary variables in annotation layers by means of guards.
Our proposal only considers how to add annotations to existing initialisation and arc expressions, and thereby only considers how to annotate existing tokens.
However, it might be useful also to be able to add auxiliary net structure to
the annotation layers. As an example, a place could be added only to the annotation layer with a token holding a counter with the number of occurrences of a transition. Allowing such additional net structure in the annotation layers would make it possible to take advantage of the powerfulness of the graphical notation of CP-nets when encoding the logics of annotations.
Another issue that should be addressed in the future is whether or not the translation from a CP-net and the corresponding annotation layer to the matching CP-net should be conducted in a tool. The author of this disserta-tion currently believes that the transladisserta-tion should be conducted implicitly by modifying the simulator in the tool. The simulator should be modified so that the annotations are handled explicitly by the simulation tool instead of being integrated into the CP-net. In other words, the translation should only be used for defining the semantics of annotations and should not be applied to every annotated CP-net. However, this should be investigated in the future.
Finally, the similarities between the monitoring framework and annotations could be considered. Both monitors and annotations aim at separating code and information from a CP-net. Annotation layers have the advantage compared to monitors, that there is a tight coupling between the annotations and the graphics of the CP-net while still easily being able to disable annotations when needed. Therefore, it should be investigated how these advantages can be used for integrating monitors into annotation layers.
Part II
Papers
39