Annotated Simulation-Based Dataset of Interaction Events in Underground Mixed Traffic: Automated Vehicle Fleets Operating Alongside a Misbehaving Human-Controlled Vehicle

Dataset and generated maps The dataset contains 725 036 interaction events generated from 3 060 unique traffic simulation scenarios involving fleets of automated vehicles (AVs) operating alongside a manually operated vehicle (MV) in underground mining environments. The scenarios are simulated across 20 mine map layouts, 3 grid maps, 10 position configurations per map, and 2 fleet sizes. Interaction events are split into 3 CSV files by event category: 451 847 - Event category 1: Proactive Slowdown events, triggered when a path overlap with an MV is detected 271 272 - Event category 2: Response to Priority Violation events 1 917 - Event category 3: Rerouting events, triggered when the MV moves slowly or remains idle on the same path Each row in a CSV file represents a single AV–MV interaction event. Columns include dynamic features (event descriptors), static features (map-specific parameters, number of AVs, and number of path overlaps), and computed performance indicators, such as Path Overlap Density (POD)-based metrics, path length, and corresponding target performance values (collision occurrence, and average AV speed or AV travel time). Encoded values: Event category (1 = Path overlap with an MV; 2 = Priority violation; 3 = Slow MV ahead) AV strategy: In event category 1 (0 = Slowdown No; 2 = Slowdown Yes, reduce maximum speed by 2 m/s) In event category 2 (0 = Ignore MV; 1 = Change of Priorities; 2 = Stop) In event category 3 (0 = Rerouting No; 1 = Rerouting Yes) Distance to other future PO start (MV): (-2) = MV is idle (-1) = no other future PO for MV Otherwise, a real value in meters Collision occurrence (0 = No collision, 1 = Collision) Features The dataset contains 24 features (excluding the columns with Event ID, Event category, and target labels), covering the following: AV strategy Dynamic features: Speed current (AV), m/sSpeed current (MV), m/sDistance to PO start (AV), mDistance to PO end (AV), mDistance to other future PO start (MV), mNo. of path points to PO start (MV)No. of path points to PO end (MV)POD at current path point (AV) POD at current path point (MV) POD score for PO with MV POD future (next 20 m) with MV POD future (next 50 m) with MV POD future (next 100 m) with MV POD future (next 200 m) with MV POD score for PO with all POD future (next 20 m) with all POD future (next 50 m) with all POD future (next 100 m) with all POD future (next 200 m) with all Static features: No. of Path Overlaps No. of AVs Map connectivity level No. of OPs Maps: The dataset also includes a ZIP folder containing 3 generated grid maps in PNG format, in addition to the mine map layouts available at https://doi.org/10.5281/zenodo.18375934 Tools used to generate the data A custom map generator for underground road networks ORU Coordination Framework, a centralized planning and coordination platform for multi-vehicle fleets