|Authors||T. Ma, S. Ali and T. Yue|
|Title||Fragility-Oriented Testing with Model Execution and Reinforcement Learning|
|Publication Type||Technical reports|
|Year of Publication||2017|
|Publisher||Simula Research Laboratory|
Self-healing is becoming an intrinsic feature of Cyber-Physical Systems (CPSs), which allows them to recover from faults in an autonomous manner during their operation. It becomes even more challenging when testing self-healing behaviors of CPSs in the presence of environment uncertainty. Such uncertainty makes system behaviors even more unpredictable in addition to the autonomous nature of self-healing behaviors. To this end, we propose Fragility-Oriented Testing (FOT) relying on model execution and reinforcement learning to efficiently test self-healing behaviors of a CPS in the presence of environment uncertainty. To evaluate the efficiency of FOT, we compared it with random testing (RT) and coverage-based testing (CBT). Results show that FOT significantly outperformed the two baselines for 7 out of 10 experiments in terms of fault revelation. Compared with RT and CBT, FOT respectively reduces 50% and 35% test execution time to find a fault.