The Experimental Data Sets and Appendix of the paper “Communication Topology Reconstruction for a Three-Dimensional Persistent Formation with Fault Constraints”

According to the different proportion of agent loss and link loss in faults, five experimental data sets are designed. Among them, the fault type corresponding to data set A is agent loss where the proportion of link loss in faults is 0%, the fault type of data set B, C and D is agent & link loss where the proportion of link loss in faults is 25%, 50% and 75% respectively, and the fault type of data set E is link loss where the proportion of link loss in faults is 100%.
In each experimental data set, experimental instances are generated under different number of agents, formation shape, and fault ratio. Specifically, the number of agents is 20, 30, 40, and 50 respectively, 5 different formation shapes are randomly generated in the area of 5000×5000×5000 under the same number of agents, and the fault ratio is 0.05, 0.10, 0.15, 0.20, 0.25, 0.30, 0.35, 0.40, 0.45, and 0.50 respectively. Therefore, the total number of experimental instances is 5×4×5×10 = 1000. Specifically, the fault ratio under agent loss is the ratio of the number of agents suffering agent loss to the number of all agents in the initial optimal communication topology, the fault ratio under link loss is the ratio of the number of communication links suffering link loss to the number of all communication links in the initial optimal communication topology, and the fault ratio under agent & link loss is the sum of fault ratio under agent loss and fault ratio under link loss, where the fault ratio under agent loss is equal to the fault ratio under link loss.
Each experimental instance has an identifier that includes the corresponding data set, the number of agents and the formation shape. For example, A-20-1 indicates that this experimental instance belongs to data set A, its number of agents is 20, and its formation shape is the first one. For each experimental instance, the corresponding formation communication cost of the solution obtained by each algorithm and its calculation time are listed in the Appendix.

本研究根据故障中智能体失效（agent loss）与通信链路失效（link loss）的占比差异，设计了五组实验数据集。其中，数据集A对应的故障类型为仅智能体失效，此时故障中链路失效占比为0%；数据集B、C、D的故障类型为智能体与链路联合失效，此时故障中链路失效占比分别为25%、50%与75%；数据集E的故障类型为仅链路失效，此时故障中链路失效占比为100%。
每组实验数据集下，均基于不同的智能体数量、编队构型与故障占比生成实验实例。具体而言，智能体数量分别取20、30、40与50；在相同智能体数量下，于5000×5000×5000的三维空间内随机生成5种不同编队构型；故障占比则分别设置为0.05、0.10、0.15、0.20、0.25、0.30、0.35、0.40、0.45与0.50。综上，实验实例总数量为5×4×5×10=1000。
具体而言，仅智能体失效场景下的故障占比，指发生智能体失效的智能体数量与初始最优通信拓扑（initial optimal communication topology）中总智能体数量的比值；仅链路失效场景下的故障占比，指发生链路失效的通信链路数量与初始最优通信拓扑中总通信链路数量的比值；智能体与链路联合失效场景下的故障占比，则为智能体失效占比与链路失效占比之和，且该场景下两者占比相等。
每个实验实例均配有唯一标识符，该标识符包含所属数据集、智能体数量与编队构型编号。例如，A-20-1代表该实验实例属于数据集A，智能体数量为20，编队构型为第1种。
针对每个实验实例，各算法所得解对应的编队通信代价及其计算耗时均列于附录中。