Data underlying the publication: The effect of models of fugitive behavior on police interception strategies

This repository is part of the Ph.D. thesis of Irene S. van Droffelaar, Delft University of Technology.<br><em>fug_behavior </em>contains the experiments and data files:- prep_graph_*.ipynb (e.g., [prep_graph_Manhattan.ipynb](prep_graph_Manhattan.ipynb) import the graph from OpenStreetMap and the camera data from the data folder. The output of these files is the plotted graph of the respective area and the saved graphs.- [enrich_graph_cool.ipynb](enrich_graph_cool.ipynb) and [enrich_graph_cool.ipynb](enrich_graph_cool.ipynb) import the relevant factors from OpenStreetMap and other data sources. These factors are used to create the graph attribute 'travel_time_adj' which reflects the attractiveness of a particular road segment. After, the best routes from the incident to each of the defined escape nodes are generated and saved.- [run_model_cool.py](run_model_cool.py) and [run_model_hot.py](run_model_hot.py) import the escape routes and pass them to a pyDSOL model, which adds a jitter factor to the routes. The output of these files is the resulting, final simulated escape routes.- [optimize/optimize_positions.py](optimize/optimize_positions.py) import the resulting routes and optimizes the positions of the police units. The optimized positions and the resulting interception dictionary (reflecting which routes are intercepted by the chosen strategy) are saved.- In [optimize/cross_evaluation_optimization.ipynb](optimize/cross_evaluation_optimization.ipynb), these positions and the routes are imported to evaluate their relative robustness. I.e., how well does a strategy perform on a set of routes generated using a different rationale or profile - also see the figure below. The resulting heatmaps are saved in the [optimize/](optimize/) folder.<br><em>platypus-fork</em> contains the optimization algorithm.

本代码仓库隶属于代尔夫特理工大学(Delft University of Technology)Irene S. van Droffelaar的博士学位论文。
*fug_behavior* 模块包含全部实验与数据文件：
- prep_graph_*.ipynb（例如 prep_graph_Manhattan.ipynb）：从开放街道地图(OpenStreetMap)导入路网数据，并从数据文件夹中读取摄像头数据，此类脚本的输出为对应区域的可视化路网与持久化存储的路网文件。
- enrich_graph_cool.ipynb（共两份，原文存在重复粘贴）：从开放街道地图(OpenStreetMap)及其他数据源导入相关影响因子，用于构建路网属性「travel_time_adj」（调整出行时间），该属性可反映特定路段的通行吸引力。完成因子导入后，脚本将生成并保存从事发地点至所有预设疏散节点的最优路径。
- run_model_cool.py 与 run_model_hot.py：导入已生成的疏散路径并将其传入pyDSOL模型，该模型会为路径添加扰动因子，最终输出模拟得到的最终疏散路径。
- optimize/optimize_positions.py：导入生成的路径结果，对警力部署位置进行优化，并保存优化后的部署位置与拦截字典——该字典用于反映选定策略可拦截的路径集合。
- 在 optimize/cross_evaluation_optimization.ipynb 中，将上述部署位置与路径导入以评估策略的相对鲁棒性，即某一部署策略在基于不同逻辑或参数集生成的路径集合上的表现效果，详见下图。生成的热力图将保存至 optimize/ 文件夹下。
*platypus-fork* 模块包含本次实验所用的优化算法。