Introducing the transitional autonomous vehicle lane-changing dataset: Empirical Experiments

A research team from North Carolina State University has introduced a crucial new dataset for autonomous vehicle development called the North Carolina Transitional Autonomous Vehicle Lane-Changing (NC-tALC) Dataset. This high-fidelity resource specifically targets the complex interactions of 'transitional autonomous vehicles' (tAVs)—systems operating between basic driver assistance (SAE Level 2) and full autonomy. The dataset's core value lies in its controlled, empirical capture of two critical scenarios: a tAV performing lane changes in traffic with adaptive cruise control (ACC) vehicles, and tAVs reacting as followers when another vehicle cuts into their lane.

The dataset comprises 152 meticulously recorded trials, split into 72 lane-changing executions and 80 follower-response maneuvers. Each trial is sampled at 20 Hz with centimeter-level accuracy using Real-Time Kinematic GPS (RTK-GPS), providing an unprecedented view of vehicle dynamics. This level of detail is essential because mandatory lane changes, like those near highway exits, are high-risk maneuvers where the interaction patterns between semi-autonomous systems and human-driven vehicles are not fully understood. The scarcity of such real-world interaction data has been a major bottleneck for researchers and engineers.

By providing a rigorous empirical foundation, the NC-tALC dataset allows developers to move beyond simulations and test how their tAV decision-making algorithms perform in realistic, interactive scenarios. It enables the analysis of traffic stability implications and safety-critical response dynamics. This publicly available dataset on arXiv is poised to accelerate research into making semi-autonomous lane-changing behavior safer, more predictable, and better integrated into mixed traffic environments, ultimately paving the way for higher levels of vehicle automation.