In recent years, end-to-end autonomous driving frameworks have been shown to not only enhance perception performance but also improve planning capabilities. However, most previous end-to-end autonomous driving frameworks have primarily focused on enhancing environment perception while neglecting the learning of autonomous vehicle planning intent. Within the end-to-end framework, this paper proposes a method termed NTT, which obtains explicit planning intent through the navigation path. NTT first generates the future target point for the autonomous vehicle based on the navigation path, thereby enhancing planning performance within the end-to-end framework. On one hand, the generation of the target point allows the autonomous vehicle to learn explicit intention from the navigation path, enhancing the practicality of planning. On the other hand, planning trajectory generated based on the target point can adapt more flexibly to environmental changes, thus effectively improving planning safety. We achieved excellent planning performance on the widely used nuScenes dataset and validated the effectiveness of our method through ablation experiments.
翻译:暂无翻译