Dataset supporting "Heavy-Duty Agricultural Truck Navigation: Geometry-Dependent Planner–Controller Trade-offs in a Human-Referenced ROS 2 Navigation2 Benchmark"

This dataset contains the experimental recordings supporting the study “Heavy-Duty Agricultural Truck Navigation: Geometry-Dependent Planner–Controller Trade-offs in a Human-Referenced ROS 2 Navigation2 Benchmark.” The dataset comprises 227 ROS 2 recordings: 212 simulation runs, including 200 nominal benchmark runs and 12 separate Scenario 2 RPP diagnostic runs; 12 professional-driver recordings used to construct the human-reference trajectories; and 3 full-scale field obstacle-avoidance recordings corresponding to the human driver, NavFn–MPPI, and SMAC–MPPI conditions. The simulation benchmark evaluates NavFn and SMAC global planners combined with MPPI and Regulated Pure Pursuit (RPP) controllers on a heavy-duty Ackermann-steered agricultural truck across straight and curved geometries. Only recordings used in the analyses reported in the manuscript and Supporting Information are included. Failed development attempts, discarded trials, obsolete recordings, and unrelated experimental data are excluded. The dataset is distributed as losslessly compressed archives organized by experimental condition. A public manifest maps every recording to its scenario, target speed, planner, controller, run number, archive member, original file size, and SHA-256 checksum, allowing verification of the recovered MCAP files against the source recordings used in the analysis.