In this project, automating timed automata design process is studied from different perspectives that are aligned with the project objectives.
We developed a new method for construction of TA models and generation of specification formula from descriptions given in structured natural language (WP1). The method is implemented as a tool named ATAC. The tool, its manual, and a white paper explaining the theory behind it are available on the project web-page.
For control synthesis for parametric timed automaton (PTA) problem, we showed that the undecidability results on the parameter synthesis problem applies to this problem (WP2). For this reason, we formulated the synthesis problem for PTA restricted to bounded integer parameters under reachability specifications for deterministic systems (WP2), and developed an efficient synthesis algorithm using iterative feasibility analysis (WP3). We extended the algorithm to synthesize optimal control strategy and parameter valuation pair (WP4). We also extended it for non-deterministic systems under unavoidability specifications (guaranteed reachability for non-determinism) (WP3-4).
We studied the synthesis problem for safety specifications for a special type of timed automata embedding the controller synthesis into the parameter synthesis problem. We designed a parametric TA of this type modelling an adaptive cruise control system (ACC), and presented efficient synthesis methods using the monotonicity properties of this model (WP3-4-5).
The controller and parameter synthesis methods intend to assist the design when a partial model is developed. On the other hand, in general, a designer makes some design choices regarding the uncertainties to produce a complete model. This model needs to be updated if it does not satisfy its specification. At this end, we studied tuning constraints of a TA for reachability, data-driven controller synthesis for safety, and finally, data-driven system repair for safety (WP3-4-5). Each of these problems is closely related to the control of parametric timed automata problem since each of them can be cast as this problem by parametrizing the considered TA. For reachability, we developed a method to find a minimal amount of change in the timing constraints(in collaboration with Prof. Cerna, initiated during the secondment). In data-driven system repair approach, we find the causes of undesired events as a temporal logic formula from system traces, and repair the system in an automated way to avoid the satisfaction of the formula. In data-driven controller synthesis, we synthesize controllable formulas that explain events leading to desired behavior and synthesize controllers from such formulas.
Finally, we developed a novel clock reduction method that reduces the number of clocks while preserving semantics and design choices to reduce the computational complexity.
At the end of the project, 3 conference papers have been published, 2 journal papers are under review, 1 conference paper is under review, another one is written and will be sent in October, 2020. A web-page for the project was established. The key-results, the progress reports and the developed tools (available for download) are shared through the web-page. Two articles are written for the general public.
To communicate the results with different audiences, I had meetings with SMEs from METU TechnoPark, gave informal talks to K-12 students, and had a stand for the project at Science is Wonderful, 2019 (Brussels).
