The shooting and bouncing ray approach is an important method used in ray-tracing simulations to estimate the propagation of radio waves. However, in this method, unnecessary rays, which are useless for follow-up and consume a large amount of simulation resources, may emerge. This study proposes an artificial intelligence-based acceleration solution, unlike geometry-based approaches in the literature. The proposed approach focuses on ray characteristics, not ambient geometry. An analysis of ray characteristics is carried out by trained simulation-decision engines at run-time. The decision engine training was carried out with datasets obtained from traditional shooting and bouncing simulations that use different parameters. The success of the proposed method was tested for three different real environments containing line-of-sight and non-line-of-sight receiver points, where the received signal strength measurements had been made earlier. For better training and more detailed analysis, a total of 63 simulations using different parameters were made in both training data and test simulations. The obtained results show the success of the proposed method.

射击和弹跳射线方法是射线追踪模拟中用于估计无线电波传播的重要方法。但是，在这种方法中，可能会出现不必要的光线，这些光线对后续无用，并且会消耗大量的仿真资源。与文献中基于几何的方法不同，本研究提出了一种基于人工智能的加速解决方案。所提出的方法侧重于光线特性，而不是环境几何。光线特性的分析由经过训练的模拟决策引擎在运行时执行。决策引擎训练是使用从使用不同参数的传统射击和弹跳模拟获得的数据集进行的。在包含视距和非视距接收器点的三种不同实际环境中测试了所提出方法的成功，其中接收信号强度测量已较早进行。为了更好的训练和更详细的分析，在训练数据和测试模拟中总共进行了 63 次使用不同参数的模拟。获得的结果表明所提出的方法是成功的。