Scoring autonomous vehicle trajectory using reasonable crowd data

Embodiments are disclosed for scoring one or more trajectories of a vehicle through a given traffic scenario using a machine learning model that predicts rationality scores for the trajectories. In an embodiment, a rendering of two or more vehicle trajectories through the same or different traffic scenarios is presented to a human annotator referred to as a "reasonable population". The annotators are required to indicate that the annotators preference one trajectory rather than other trajectory (s). A machine learning model is trained using inputs collected from human annotators to predict rationality scores for one or more trajectories for a given traffic scene. The predicted trajectories may be used to rank the trajectories based on scores of the trajectories generated by a route planner, compare AV software stacks, or may be used by any other application that may benefit from machine learning models that score vehicle trajectories. 公开了用于使用机器学习模型来对运载工具通过给定交通情景的一个或多个轨迹进行评分的实施例，该机器学习模型预测这些轨迹的合理性得分。在实施例中，向被称为"