%0 Dataset
%T Dataset for Cross-Scenario Collaborative Interception by Unmanned Swarms
%J National Cryosphere Desert Data Center
%I National Cryosphere Desert Data Center(www.ncdc.ac.cn)
%U http://www.ncdc.ac.cn/portal/metadata/3f9e8eaf-d0d7-4bfd-8545-67972b256f8d
%W NCDC
%A yu wen wu
%K Coordinated interception;Multi-agent reinforcement learning
%X This dataset was established in July 2025 and generated using a Python simulation platform. It is primarily designed for the large-scale unmanned swarm cooperative interception problem in complex dynamic scenarios, aiming to support performance analysis and validation of multi-agent reinforcement learning, dynamic target grouping, graph attention relationship modeling, and cross‑scenario generalization algorithms. The dataset is generated from a 2D swarm interception simulation environment, which constructs adversarial interaction scenarios between a friendly missile swarm and an enemy UAV swarm. The training scenarios mainly involve 20 friendly missiles intercepting 10 enemy UAVs, with extensions to different scales such as 15vs10 and 10vs10, as well as test configurations including in‑training policies, out‑of‑training policies, different enemy speeds, and different enemy formations. The dataset contains a single TXT text file, namely r_replay_bufer.txt. This file records the reinforcement learning elements—states, actions, rewards, etc.—during the unmanned swarm's mission execution. It contains four columns of data, namely the swarm state, individual swarm actions, the team reward function value, the next swarm state, and a flag indicating whether the current task is completed (done). The data content includes agent states, action commands, reward signals, task termination flags, dynamic target grouping results, enemy subgroup characteristics, graph attention weights, and evaluation metrics such as task success rate, average number of kills, and average number of steps. This dataset can effectively support training of cooperative interception policies for unmanned swarms, analysis of target assignment mechanisms, and testing of cross‑scenario generalization performance.