SUMMARY: Unmanned aerial vehicle (UAV) was introduced for nondeterministic traffic monitoring, and a real-time UAV cruise route planning approach was proposed for road segment surveillance. First, critical road segments are defined so as to identify the visiting and unvisited road segments. Then, a UAV cruise route optimization model is established. Next, a decomposition-based multi-objective evolutionary algorithm (DMEA) is proposed. Furthermore, a case study with two scenarios and algorithm sensitivity analysis are conducted. The analysis result shows that DMEA outperforms other two commonly used algorithms in terms of calculation time and solution quality. Finally, conclusions and recommendations on UAV-based traffic monitoring are presented.
Traffic conflict technology (TCT) is widely used to assess the safety of work zones. The current TCT is temporal and (or) spatial proximity defined based, which can only detect two-vehicle or multi-vehicle conflicts, and is not competent for single-vehicle conflicts. However, single-vehicle accidents in work zones are also severe. This study proposes a detection method for all types of traffic conflicts in work zones. Based on vehicle micro-behavior data, evasive behavior is detected by automatic segmentation, Support Vector Machine (SVM)-based behavior identification, and threshold-based judgment methods. Two-vehicle or multi-vehicle conflicts are detected by classical proximity defined-based method, i.e., the surrogate safety assessment model (SSAM). By comparing the analysis results of the evasive behavior with the one of SSAM, single-vehicle conflicts can be detected. Taking a practical work zone as an example, the effectiveness of this method in detecting all types of traffic conflicts in work zones is verified. The single-vehicle conflict can be subdivided into 10 types according to basic behavior types, such as straight-line driving and decelerating. The two or multi-vehicle conflicts can be subdivided into three types, such as rear-end conflict. The example also verifies the applicability of this method under different traffic volume scenarios.