Research Focus:
- Search-based software engineering and testing of complex systems
- Automated test generation and validation techniques for cyber-physical systems (CPS)
- Application of machine learning and artificial intelligence in software testing
- Multi-objective and multi-requirement approaches to CPS testing
- Exploration of Reinforcement Learning (RL) for test generation and validation of CPS
Key Contributions:
- Used WOGAN (Wasserstein Generative Adversarial Network) algorithm for CPS testing, employing a modified fitness function with multiple objectives
- Contributed in the development of the Explicit Output Coverage (EOC) algorithm
- Created the WOGAN-UAV tool for adaptive test generation of unmanned aerial vehicles
- Participated in the SBST (Search-Based Software Testing) 2023 and SBFT (Search-Based and Fuzz Testing) 2024 tool competitions, demonstrating the effectiveness of WOGAN-based approaches
- Helping develop STGEM (System Testing Using Generative Models) to incorporate reinforcement learning techniques
Methodologies:
- Black-box testing approaches
- Online learning and adaptive test generation
- Utilization of fitness functions and coverage metrics to guide test generation
- Signal Temporal Logic (STL) for specifying and monitoring temporal properties
- Multi-objective optimization techniques
Applications:
- Testing lane keeping assist systems in autonomous vehicles
- Generating obstacle scenarios for UAV collision avoidance testing
- Falsification of safety requirements in CPS
The research aims to improve the efficiency and effectiveness of testing safety-critical cyber-physical systems through advanced search-based techniques, machine learning, and formal methods, with a growing focus on handling multiple requirements and exploring reinforcement learning approaches.
