Optimizing for Ride Comfort: A Model Predictive Control Framework with Frequency-Domain Analysis of the Acceleration Sequence

Abstract

Improving ride comfort can help accelerate the adoption of autonomous vehicles (AVs). Unfortunately, very few studies directly consider comfort in the controller design and among the existing ones, most of them solely focus on instantaneous acceleration and jerk. Such an approach cannot fully capture ride comfort. In fact, the International Organization for Standardization (ISO) emphasizes that ride comfort should be evaluated based on acceleration patterns over time. To bridge this research gap, this study proposes a comfort-centric Model Predictive Control (MPC) framework that optimizes both tangential and lateral acceleration patterns for optimal 2D maneuvers, including longitudinal acceleration and steering rate. The framework is subsequently tested against turning trajectories from the Waymo Open Dataset. Results demonstrate that our approach improves ride comfort compared to the original Waymo trajectories. Here, more comfort improvement can be achieved at higher lateral acceleration, implying that the proposed MPC framework can lead to more gentle turning behaviors. These findings highlight the effectiveness of the proposed MPC framework in enhancing ride comfort.