| dc.description.abstract | A real-time system (RTS) is usually well-defined and operates based on a specific model defined during system design. However, the RTS can interact with different objects from its environment and needs to satisfy a number of user-defined constraints such as safety (defined using the probability of failure) and performance (defined using the percentage of usage). Such requirements create the necessity for the RTS to be aware of its design and execute a set of additional tasks (apart from the tasks whose order is defined by a particular scheduler during system design) in response to the events which take place in the environment.
This thesis presents the design of a situation-aware RTS which can characterize the environmental situations through monitoring the system environment, analyzing the input obtained from the environment and identifying real-world occurrences as events.
Additionally, we determine the real-time and non real-time properties associated with the events, identify the relationships involved among the events and create a knowledgebase offline which facilitates a reduced size of data for storage and processing.
We present a situation-aware task model (SATM) which efficiently maps the identified environmental events to a set of (predefined) adaptive tasks offline. This thesis also presents a validation framework which determines the user-defined safety, and performance constraints. We consider that the situation-aware RTS has two modes of operation: safety, and performance. The validation framework performs an online identification of the expected mode based on the user-defined constraints, checks whether the
RTS is operating in the correct mode or not and allows the RTS to change its operating mode (if necessary).
To demonstrate the applicability of the proposed situation-aware RTS and usability of the SATM, the experimental analysis of the thesis is performed using three case studies: an automotive system, a real-time traffic monitoring system and an unmanned aerial vehicle (UAV) system which include RTS that are in motion and static. For the automotive system case-study, the experimental results of this thesis show that we identify 17234 events in 3241 environmental situations. The system operates in performance mode in 3295 situations and in safety mode in 126 situations when the probability of failure is high. The system consists of five tasks in the performance mode and three tasks in the safety mode and the corresponding constructed SATM contains nine vertices (adaptive tasks) and 68 edges. For each case-study, the constructed SATM provides an improvement in terms of scheduling overhead (up to 21%) and adaptation time (up to 49%) with respect to existing task models task models such as generalized multiframe model (GMF), non-cyclic generalized multiframe model (NC-GMF), recurring branching (RB), recurring real-time task (RRT), and non-cyclic recurring real-time task model (NC-RRT). | en |
