Benchmark instances for the stochastic hybrid truck-drone routing problem

To evaluate the performance of the proposed model in figure 1, we generated 180 small-sized instances and 180 medium-sized instances, and 60 large-sized instances. We generated customer locations in a 1000*1000m2 square. Then randomly assigned to three customer classes. As the second-class and the third-class customers must be covered with the rendezvous locations, we created them in a way that at least one rendezvous node is in sight radius of each second-class or third-class customer. The sight radius is considered to be 100 meters for this data generation. We have defined two locations for the depot as follows: the vertex of the square <em>(0,0)</em> and the center of the square <em>(500,500)</em>. We assumed 10 meters per second for truck speed and 20 meters per second for drone speed. In this way, we created 36 various instance types and generated 10 instances for each type.

为评估图1中所提模型的性能，我们共生成180个小型实例、180个中型实例与60个大型实例。我们在1000×1000平方米的正方形区域内生成客户位置，并将客户随机划分为三类。鉴于二类与三类客户必须依托集结点完成服务覆盖，我们在生成该两类客户时，确保每一个二类或三类客户的可视范围内至少存在一个集结节点。本次数据生成过程中，设定可视半径为100米。我们为配送场站设定了两处选址：分别为该正方形区域的顶点<em>(0,0)</em>与区域中心点<em>(500,500)</em>。我们假定卡车行驶速度为10米/秒，无人机飞行速度为20米/秒。通过上述参数设置，我们共生成36种不同的实例类型，并为每类实例生成10个具体实例。