The PROBAnD Railway Crossing Specification

To better understand our design contest results (especially the final SDL specification), which have been attained with an extension and modification of the PROBAnD method, an informal description of the method's development activities and documents will be given in this document.

As an input, the requirements engineering method PROBAnD takes the problem description, which is divided into the building description and a collection of needs (c. f. Figure 1).

The building description contains a description of the building’s structure. Further, the building’s installation is depicted. From this building description, an initial control system structure can easily be derived because the control objects are most often identical to the the building’s objects or are a sub-set of these. To handle the huge number of control objects, control object types are formed, which are aggregated according to the hierarchy of the building’s objects. As requirements engineering proceeds, this control system structure can be refined, as long as the strict aggregation hierarchy of control object types is maintained.

The other part of the problem description is made up by a collection of needs. These needs informally describe the control system from the point of view of the users. The needs are split into less complex control tasks such that they can be assigned to single control object types.

As a guideline for the above steps, we suggest that the responsibilities of a control object type at a certain level should match the control tasks that can be performed at that level, which allows minimizing the flow of information. This distribution of responsibilities, which can easily be applied in the domain of building automation systems, has similarities to an organizational hierarchy. Communication between control objects is realized through the exchange of signals (which may have parameters) that are only allowed to travel along the aggregation hierarchy. This helps maintaining the easy comprehensibility of the specification, and allows the creation of documents from a set of few templates to simplify recurring activities (like specifying communication channels between control object types).

After the control object types have been identified and control tasks have been assigned, strategies for realizing the control tasks are informally given in natural language, leading to a collection of semi-formal control object types, which are expressed in HTML documents.

From these semi-formal control object types, operational control object types are specified in SDL documents, which together form an executable SDL specification.

The executable specifications are used for creating control system prototypes, with which the test of the systems becomes feasible. To ease the interaction with the system during run-time, Java panels (GUIs) can be used that allow the dynamic change of parameters and settings.

One important aspect of our process is that the development time for creating and modifying each document is recorded (for each document a "Version History" is kept), which allows measuring the overall development effort.

Design Contest Results:

• Problem Description (Version History): The description of the entities of the railway crossing (which take the part of the "building description"), and the collection of needs, are contained in a single document, the original statement of requirements of the SAM design contest. The initial problem description has been extended to show how the specification could be addapted to these new requirements. These changes are reflected in the Version Histories of the respective documents by the version comment 'CHANGE of requirements regarded' or a similar text. The effort for these modifications is listed below.

 
• Control System Structure (UML-Diagram, Version History): According to the approach for the building automation domain, the system structure has been attained very early in development by using the entities of the physical system as a starting point; e.g. the trains, the gate or the sensors. As the whole railway crossing system had to be specified, the controller part and the actual environment had to be considered. Thus, each entity is represented from the point of view of the controller (XxxCtrl) and from the point of view of the environment (EnvXxx). Note, that each sensor and actuator on the Env-side of the system communicates with the respective entity on the Ctrl-side of the system. As an extension to the PROBAnD approach, some entities in the control system (e.g. trains) are instantiated during run-time, i.e. after the system has been initialized. This is specified by a variable multiplicity of n in the documents.

 
• Control Tasks: The control tasks have been directly specified in the semi-formal control object type documents because their assignment to control object types has already been known when the tasks were specified.

 
• Semi-Formal Control Object Types:
• The core of the Env-side of the system is realized by the control object type EnvTrain (Version History). The train sensors and the signal and gate actuators merely change their states. Each EnvTrack (Version History) is responsible for creating the trains with a specified delay or upon manual request from the panel.
• The strategies for controlling the gate and the signals depending on the information from the sensors is realized on the Ctrl-side by the RWCrossingCtrl object type (Version History) together with TrackCtrl (Version History). Four strategies, as described in the HTML control object type document, are possible.

 
• Operational Control Object Types:  As an extension to the PROBAnD method, the instantiation of control object types during run-time is realized by dynamically creating SDL processes.

 
• SDL Specification (Version History, Telelogic Tau Organizer File for WIN32): The specification has been used to generate protoypes of the railway crossing system with Telelogic Tau. In order for the system to start up correctly, a file named sockadr needs to be created in the current directory. This file has to contain four lines, stating the correct hostname and port for the panel, the last two lines can be set to an arbitrary value.

For initializing the system (number of tracks, settings of distances, ...) a Java panel can be used. However, if the Java platform is not available (or difficulties in connecting the panel to the prototype persist) the simulator interface can be used to start a demo. Proceed as follows:
• start (realtime) simulation, then enter:
• set-trace 1
• go
• wait, until no more output is visible on stdout, then stop simulation (break), and enter:
• Output-To physValIn ('obj1', 'demo') ProtoCtrl

 
• Test Cases: For Testing the system behavior, each active entity of the model (trains, signals and the gate) creates a logfile in the logs-directory (for an example see here). Each of these files is named after the actual instance name of the object concatenated with 'Log' and contains information like the position of a train, the state of a signal, etc. These files have been used to create graphical outputs for each test-case, which show the behavior of the system in an easily comprehensible form (for an example see here).

 
• Java Panel: The panel can be used for dynamically interacting with the railway crossing system. After the prototype has been started, the panel can be connected to the hostname and port specified in the sockadr-file (see above) as follows:
• change to the Java-Directory, where the Java-Files are located, enter:
• java main.RWPanel obj1 <hostname> <port>

 
• Effort: The effort of each development activity is recorded in the Version Histories of each document and also in the overall Process Trace. The following table shows the effort distribution (according to the activities introduced above). The total effort for the specification of the railway crossing was 4134 min (68.9 hrs).

 

Activity	Affected Part of System	Initial Effort [min]	Effort for Modification [min]
Constrol System Structure Specification		54	0
Control Task Description + Control System Description	Env-side	493	2
	Ctrl-side	161	16
	Panel, Top-level Object Type	106	0
Control System Modeling	Env-side	1032	4
	Ctrl-side	197	54
	Panel, Top-level Object Type, Signal Definitions, Tests	2012	3
TOTAL		4055 = 67.6 hrs	79 = 1.3 hrs