SF-LIFE
收藏资源简介:
SF-LIFE是由多个研究机构联合创建的大规模模拟移动数据集,旨在为交通、移动性和机器学习研究提供无噪声、完整的多模态轨迹数据。该数据集包含3万亿条位置记录,模拟了50万个智能体在70天内的1Hz频率移动轨迹,数据融合了基于智能体的人类生活模式仿真产生的日常议程,并利用旧金山湾区9个县40多家交通机构的OpenStreetMap和GTFS数据生成详细运动轨迹,覆盖公交、铁路、自行车、汽车和步行等多种交通模式。数据集通过基于马斯洛需求层次理论的智能体行为框架生成,结合人口普查数据和真实交通基础设施,模拟了智能体的强制性活动与需求驱动活动。其核心应用领域包括交通优化、人类移动性分析、城市计算和空间数据分析,旨在克服真实世界追踪数据中的隐私、噪声和完整性限制,为算法开发和模型基准测试提供高质量的合成数据资源。
SF-LIFE is a large-scale simulated mobility dataset jointly created by multiple research institutions, aiming to provide noise-free, complete multi-modal trajectory data for transportation, mobility and machine learning research. This dataset contains 3 trillion location records, simulating 1Hz-frequency movement trajectories of 500,000 agents over 70 days. The data integrates daily agendas generated by agent-based human lifestyle pattern simulations, and generates detailed movement trajectories using OpenStreetMap and GTFS data from over 40 transportation agencies across 9 counties in the San Francisco Bay Area, covering multiple transportation modes including bus, railway, bicycle, car and walking. The dataset is generated through an agent behavior framework based on Maslow's hierarchy of needs, combining census data and real transportation infrastructure to simulate agents' mandatory activities and demand-driven activities. Its core application fields include transportation optimization, human mobility analysis, urban computing and spatial data analysis. It aims to overcome the privacy, noise and integrity limitations of real-world tracking data, providing high-quality synthetic data resources for algorithm development and model benchmarking.




