Overview of the Tasks

This section shows the functional view of the previous model represented by a hierarchy of tasks and methods which define the possible ways of solving problems included in the public transport management model. Then, a first version of the space of problem solvers used to generate the answers considers:

 

Figure 7: Part of the Tasks-Methods Hierarchy Focused on the Planning Functionality 

• To perform the classification of a line scenario an identify situation task first makes an abstraction of the basic state data sent by the information system, and then realizes the identification of the current situation using two types of rule-based methods: (i) a shallow one representing levels of alarm in a line, and (ii) a more detailed one which specifies the relevant issues in the line that determine the problematic situation.
• The prediction task can be realized with two methods:
  1. a classification one making use of a heuristic prediction model which associates the presence of relevant current events with predictable states compiled from the operators experience; and
  2. a more complex one where four different subtasks can distinguished: (i) estimation of the expected travel times in the different areas of the line, (ii) simulation of the short term evolution of the delayed vehicles, (iii) take of the scheduled behavior of the delayed vehicles from a database, and (iv) comparison of the expected and foreseen states of the delayed vehicles to infer possible service disruptions.
• The planning of the required control actions to improve a problematic situation can be performed with two methods:
  1. a plain/shallow one which provides a control plan expressed in general terms just to give the guidelines of the strategy to be followed. It is supported by a knowledge base of rules relating possible problematic situations in a line with high level control plans.
  2. a hierarchically structured one which builds the control plans following an establish and refine strategy supported by the organized set of specialists in different aspects of the configuration of control plans. This method performs a hierarchical search based on a loop of three stages, performed by basic tasks, up to the final specification of the control plan. These subtasks are the following: (i) the specialist analyzes if the input satisfies its applicability conditions, (ii) the refinement knowledge of the specialist is applied to identify the most adequate specification of the current control plan, (iii) the stages of the plan selected in the previous phase are analyzed to identify those whose performance requires the participation of additional specialists.

Finally, the reflective knowledge used to dynamically select the appropriate structure of tasks to solve each particular problem is distributed among the three main previous activities, i.e. classification, prediction and planning. The reasoning is performed through a selection task supported by rules relating interaction modes chosen by the user with the corresponding problem solving method.