Many application domains like office automation, planning, medical diagnosis, manufacturing systems, etc., require interactions between humans and systems, and among humans for conducting day-to-day work. Very often, problems arise such as ineffective communication among participants working on the tasks, lack of resources or data, lack of monitoring of the task's progress, lack of handling tasks when exceptional cases arise, and so on. Consequently, the tasks may be finished in an untimely and error-prone manner....[ Read more ]

Many application domains like office automation, planning, medical diagnosis, manufacturing systems, etc., require interactions between humans and systems, and among humans for conducting day-to-day work. Very often, problems arise such as ineffective communication among participants working on the tasks, lack of resources or data, lack of monitoring of the task's progress, lack of handling tasks when exceptional cases arise, and so on. Consequently, the tasks may be finished in an untimely and error-prone manner.

In this thesis, we developed an event-driven activity execution model for an Activity Management System (AMS) which is a software system facilitating the specification, maintenance, and execution of activities. The objective is to develop a a methodology and framework by which the tasks (activities in general) can be executed and completed in an effective, timely, and harmonious manner, Different types of dependencies (data, control, temporal, external) are modeled by a uniform framework of events. Based on the events that occur, the relevant ECA (event-condition-action) rules are triggered guiding the execution of an activity. The execution model consists of two main modules, namely, the activity generator and activity coordinator. The activity generator is mainly responsible for generating the run-time coordination plan in terms of events and their interactions as specified by ECA rules prior to activity execution. The activity coordinator is mainly responsible for controlling and monitoring the execution of an activity based on the events raised and their corresponding ECA rule triggerings. We also propose a framework to handle dependencies among concurrently executing activities. By defining different types of dependencies (resource sharing, PSA assignment, task specific constraints), we detect dependencies that could cause potential conflicts and specify these conflicts by means of conflict detecting rules. To assure every activity is executed and is finished in one of the pre-specified alternative execution plans, a conservative scheduling approach is applied such that only conflict-free tasks belonging to any of the concurrently executing activities are initiated to execute. A prototype is implemented for a single activity execution to illustrate the framework presented in this thesis.