Nowadays cyber-physical systems became widespread in civil, industrial and military applications. Their most prominent appearance is most relevant in the spread of robotic systems in automatization of different conventional tasks -- like cooperating robots in manufacturing plants or autonomous vehicles in traffic. Cyber-physical systems -- and especially robotic systems -- are usually accompanied with artificial intelligence-based software designed to provide some degree of autonomous operation.
The mission goal and the environment of cyber-physical systems are usually complex-structured. Moreover, these systems are usually associated with critical missions posing potential damage to property and human health through their actions. The verification and validation of these systems are essential, but a typically difficult and resource consuming task.
This thesis introduces a testing framework aiming to assist the system-level verification and validation of autonomous systems. Autonomous systems are usually accompanied by a detailed specification, which this framework uses for combined contextual and temporal evaluation of the system. The framework also focuses on simulation-based verification and is designed to interface with a widespread robotic simulator (OSRF Gazebo). The framework is also capable of generating new test cases based on contextual description which can be used to feed the simulated environment with diverse layouts.
The applicability of this framework is presented through two case-studies of different complexity. The first case study is a trivial usage example through an object detection scenario. The next case-study is the verification of a complex scenario: a tram verified in a simulated urban environment.
Future development of the framework can focus on refining the test generation based on the evaluation feedback of the monitoring of a scenario, integrating with widespread automated driving systems (like Autoware), optimizing the runtime-monitor performance to allow operation on resource-limited systems, and enhancing the usability of the framework through model-based tools.