Real-world deployment of trajectory smoothing with traffic-aware adaptive cruise control for automated vehicle speed adjustment

This paper introduces a Traffic-Aware Adaptive Cruise Control (ACC) system for production vehicles, enhanced with downstream traffic data. A machine learning-based speed prediction model is trained using this data, incorporating both the vehicle's current speed and the speed and position of the leading vehicle to improve prediction accuracy. The predicted speed profile is then refined using scaled constant averaging over predefined intervals, reducing the frequency of ACC speed adjustments to every 5 s. This strategy helps ensure smoother and more stable speed behavior, as frequent speed changes are undesirable in ACC systems. While the vehicle still operates using the default ACC system, the set speed is dynamically adjusted based on predicted future speeds, with the goal of minimizing unnecessary speed fluctuations. Instead of accelerating to an unnecessarily higher set speed when there is no leading vehicle or mimicking the exact trajectory of the leader, the system generates an equivalent, steady trajectory that avoids frequent acceleration or deceleration. Validation along the test route yielded a mean absolute error of less than 1 mph in speed prediction. Additionally, energy consumption and trajectory analyses demonstrated substantial improvements with the Traffic-Aware ACC compared to the default ACC system. Numerical simulations further confirmed the system's effectiveness in mitigating traffic waves caused by leading vehicles, emphasizing its potential to optimize real-world traffic flow and enhance energy efficiency.