DESSERT has applied rule based reasoning, case based reasoning, and constraint handling, making use of code developed within the project as well existing public and commercial software. Service resource allocation, and particularly resource configuration, has provided a rich domain in which valuable experience has been gained in the effective representation of knowledge in the form of production rules. Constraint satisfaction and propagation techniques have been convincingly applied to scheduling and configuration problems, where significant technical problems have been overcome. Additionally, problems without a strong nor well understood underlying model have been tackled using a case based approach, a technique which fits well into the decision support paradigm.
Mathematical optimisation methods have been applied to a range of configuration and scheduling problems. Standard linear programming and mixed integer and linear programming techniques have been used to optimally solve selection of telecommunications channels across multi-provider networks. Additionally, dedicated algorithms were implemented to generate network topologies and others designed to provide job schedules according to the complex interplay of temporal constraints and optimisation criteria.
Several problem domain areas were demonstrated to be amenable to a method of trade-off visualisation using multi-criteria decision making combined with fuzzy set theory. This allows sets of alternative problem solutions to be explored whilst varying the relative importance of the decision criteria in order that the most appropriate solution may be selected in an informed manner.
Operations research and knowledge based techniques tend to have complementary strengths and weaknesses. DESSERT has experimented with the co-operation of operations research (OR) and knowledge based system (KBS) in order to solve complex problems in a flexible, efficient and accurate way.
The prototype Resource Scheduling DSS demonstrates convincing co-operation between mathematical optimisation and symbolic reasoning. Optimisation capabilities have been achieved in this DSS demonstrator by the use of OR techniques, whereas, flexibility, reactivity and openness have been achieved by the use of KBS techniques An initial schedule is derived by KBS methods and then optimised by OR methods. In addition re-scheduling is managed by knowledge based symbolic reasoning, focusing on the most critically important operations, in order to avoid further computational overhead.
In the prototype Generation and Selection of Alternative Configurations DSS, the Transit Network Configuration tool (TNC) demonstrates two approaches for combining OR and KBS techniques. The first version of the TNC allows the user to choose between generation of a configuration using either linear programming techniques or by rule based heuristics. By contrast the TNC version two uses heuristics to re-form the configuration problems so that they are suitable for solution by linear programming.
DESSERT has demonstrated that OR and KBS may be appropriately combined in a number of ways: both pre and post processing of a problem by symbolic KBS techniques for OR methods; tuning of OR parameters by KBS methods; controlling OR tools with KBS tools; and also treating OR and KBS as alternative methods for the same problem.
The mechanism upon which multiple technique inter working has been based is the multi-layer Blackboard. This model of problem solving control was developed in the 1980s to provide a co-operative integration structure for knowledge based systems. In DESSERT the Blackboard architecture has been successfully used in all three prototype DSSs and in addition it has proved to be a sound integration mechanism for DSS built from the toolkit. The project has extended and enhanced its Blackboard in each application. A version of the Blackboard has been added to the DESSERT toolkit, which is automatically used when building any new DSS.
In order to develop the three prototype service management DSS major effort was directed towards studying appropriate parts of the service provisioning domain. As a result significant information and knowledge was captured. This was structured and managed using object based information modelling techniques and in some cases it was represented in other ways such as production rules or case based reasoning cases. Finally all the information modelling work was harmonised into a single model and compared with relevant models form standards bodies and relevant RACE projects.
Central to the decision support paradigm is co-operation between system and the user to jointly solve problems. This puts great importance on the point where the user and system interact. DESSERT has recognised that the quality of user interfaces is paramount. The prototype DSS and toolkit employ varied visualisation techniques to meaningfully represent information allowing both detail and context to be understood. Windowing graphical user interfaces have been used throughout and visualisation has been supported through topological and topographical maps, hierarchical tree diagrams and temporal sequence graphs.
Overall DESSERT has applied a wide range of advanced information processing techniques to decision support for service management. The techniques employed are representative of the best well established methods, some are new and experimental, others are state of the art. The project has demonstrated how heterogeneous technologies can be combined into powerful unified decision support systems.