Generated Dataset for Social Cost Function Learning

该数据集包含了50万个样本，旨在训练网络识别机器人导航中的社交场景。数据集涵盖了仅包含队列、仅包含群体以及混合场景的多种变化。生成的代价地图成功率超过95%。然而，该数据集的局限性在于仅包含直线队列，并且在拥挤环境中的识别存在问题。规模上，训练样本量达到了50万个。任务则是训练网络识别队列和社交群体互动，以便在机器人导航中进行路径规划。

This dataset comprises 500,000 samples, developed for training neural networks to recognize social scenarios during robot navigation. It encompasses various scenario configurations: pure queue-only scenarios, pure group-only scenarios, and hybrid mixed scenarios. The success rate of the generated cost maps surpasses 95%. However, this dataset has inherent limitations: it exclusively features linear queues, and encounters recognition difficulties in densely crowded environments. In terms of scale, the total number of training samples reaches 500,000. The core task of this dataset is to train neural networks to identify queues and social group interactions, so as to facilitate path planning in robot navigation.