Abstract
To ensure reliability and comfort in autonomous driving systems, vehicle motion must feel natural to occupants. Emulating strategies used by skilled drivers—such as coupling longitudinal (speed) and lateral (path) control—can enhance perceived trust and comfort. Building on this idea, we have developed a trajectory and speed planning framework that incorporates human-like motion characteristics, including jerk minimization and coupled control strategies. In prior work, we proposed an analytical method for lane changes and speed regulation on straight roads, using time-polynomial representations to avoid iterative nonlinear optimization. This approach enables real-time applicability and transparent control logic. However, we identified a ride comfort issue: abrupt changes in longitudinal acceleration when lateral acceleration reverses sign. To address this, we propose a human-inspired solution that continuously adjusts the coupling gain based on vehicle state in the lateral acceleration phase plane, rather than switching it discontinuously. Simulations demonstrate that this method significantly improves jerk and acceleration profiles, yielding smoother and more natural motion. We also visualize the tradeoff between gain transition time and comfort improvement, offering design insights for practical implementation. While effective, the analytical formulation requires high-order polynomials—eighth-order for longitudinal and ninth-order for lateral position—even for simple maneuvers like lane changes. This highlights a challenge for extending the method to more complex road geometries, such as curved or composite roads. We conclude by discussing future directions to address this limitation.
