Context:

Feasible test sequences generation is a key step in protocol conformance testing based on the Extended Finite State Machine (EFSM) model. To guarantee the feasibility of generated test sequences, transition executability analysis (TEA) technique is widely applied in automatic test derivation. However, the TEA method often suffers from the famous state explosion problem, which has become a major obstacle to its efficient application.

Objective:

In order to mitigate this issue, this paper proposed a novel heuristic TEA method (MTEA) that uses Monte Carlo tree search (MCTS) to guide the TEA tree expansion for efficiently deriving feasible test sequences.

Method:

The approach first provides a framework to apply the MCTS algorithm based on multiple decision subtrees, in the context of test sequence generation for EFSM-specified systems, to more efficiently expanding the TEA tree with huge state space, and thus alleviating the problem of state explosion. To achieve this, we then design a reward function to calculate the fitness of nodes currently being expanded in the TEA tree and heuristically direct the search towards a near-optimal solution. Next, an adaptive reduction mechanism of search budget is also introduced to accelerate the convergence of the analysis. Finally, a MTEA-based algorithm for automatically generating feasible test sequences is presented under a specific transition coverage criterion.

Results:

A detailed case study on 6 popular EFSMs was carried out to evaluate the effectiveness and efficiency of our method. Experimental results show that the MTEA significantly outperforms Breadth-First-Search based TEA method (BTEA) and the standard MCTS-based method (SMCTS), regarding time and space performance. Compared with the BTEA, SMCTS and random TEA method (RTEA), the success rate of test generation of MTEA (98.14% on average) is approximately 2, 1.85 and 3 times higher, respectively. For successful test derivation, MTEA only needs to explore on average 9.95% of the nodes and consume on average 61.68% of the runtime of the BTEA method.

Conclusion:

The experiments illustrate the promise of our approach for alleviating the state explosion problem in test generation for EFSM-specified systems.

中文翻译：

---

使用蒙特卡洛树搜索从EFSM模型生成可行的协议测试序列

语境：

可行的测试序列生成是基于扩展有限状态机（EFSM）模型的协议一致性测试中的关键步骤。为了保证生成测试序列的可行性，过渡可执行性分析（TEA）技术被广泛应用于自动测试推导中。然而，TEA方法经常遭受著名的状态爆炸问题，这已成为其有效应用的主要障碍。

客观的：

为了缓解这个问题，本文提出了一种新颖的启发式TEA方法（MTEA），该方法使用蒙特卡罗树搜索（MCTS）来指导TEA树扩展以有效地得出可行的测试序列。

方法：

该方法首先提供了一个框架，用于在针对EFSM的系统的测试序列生成的背景下，基于多个决策子树的MCTS算法可以更有效地扩展具有巨大状态空间的TEA树，从而缓解状态爆炸问题。为了实现这一点，我们然后设计一个奖励函数来计算当前在TEA树中正在扩展的节点的适应度，并启发式地将搜索导向接近最佳的解决方案。接下来，还引入了搜索预算的自适应减少机制，以加快分析的收敛速度。最后，在特定的过渡覆盖标准下，提出了一种基于MTEA的算法，用于自动生成可行的测试序列。

结果：

对6种流行的EFSM进行了详细的案例研究，以评估我们方法的有效性和效率。实验结果表明，就时间和空间性能而言，MTEA明显优于基于广度优先搜索的TEA方法（BTEA）和基于MCTS的标准方法（SMCTS）。与BTEA，SMCTS和随机TEA方法（RTEA）相比，MTEA的测试成功率（平均98.14％）分别高出约2、1.85和3倍。对于成功的测试派生，MTEA仅需要探索平均9.95％的节点，并且平均消耗BTEA方法的运行时间的61.68％。

结论：

实验表明，我们的方法有望缓解EFSM特定系统的测试生成中的状态爆炸问题。