Intelligent software debugging: A reinforcement learning approach for detecting the shortest crashing scenarios

作者：

Highlights：

• Reinforcement learning in automated testing to find crash scenarios.

• Model-free on-policy and model-based planning-capable RL agents.

• Detected Goal Catalyst as an heuristic approach.

• Results shows that Detected Goal Catalyst increases performance.

• Combination of Reinforcement learning and Recursive Delta Debugging.

摘要

•Reinforcement learning in automated testing to find crash scenarios.•Model-free on-policy and model-based planning-capable RL agents.•Detected Goal Catalyst as an heuristic approach.•Results shows that Detected Goal Catalyst increases performance.•Combination of Reinforcement learning and Recursive Delta Debugging.

论文关键词：Reinforcement learning in automated bug detection,Exploring crashes by SARSA,Exploring crashes by prioritized sweeping,Delta debugging,Detected goal catalyst

论文评审过程：Received 19 July 2021, Revised 15 February 2022, Accepted 21 February 2022, Available online 7 March 2022, Version of Record 18 March 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116722