Motions Speak Louder Than Words: Designing Human-like AV Speed Profiles at Crosswalks with a Bayesian Optimisation–Driven Adaptive Experiment
Abstract
Autonomous vehicles (AVs) must communicate intent to pedestrians in mixed traffic to ensure safe interaction, and vehicle deceleration motion is a promising implicit communication channel. Designing acceptable AV deceleration remains difficult because it is continuous with interdependent kinematics, making fixed‑design experiments inefficient. We therefore parameterised deceleration using a minimum‑jerk model (MJM), which generates smooth speed profiles from tuneable parameters. We will conduct an adaptive experiment in an immersive pedestrian simulator, where Bayesian optimisation sequentially proposes AV deceleration profiles expected to receive higher pedestrian ratings for safety and efficiency, based on previous ratings. Preliminary results from fitting MJM to 62 naturalistic human profiles indicate that the MJM reproduces the overall human profile shape (median RMSE = 0.88 m/s) with only 2 parameters (i.e., stopping distance, duration), supporting its use as a profile generator for the experiment. This work can contribute a human-in-the-loop framework for designing AV motion as an implicit interface.