ActiveAD is a planning-oriented active learning method designed to improve sample and label efficiency in end-to-end autonomous driving (E2E-AD). The method addresses the challenge of high-quality labeled data scarcity and the long-tailed distribution of data in AD tasks. ActiveAD introduces novel diversity and uncertainty metrics based on planning-oriented philosophy to select the most useful samples for annotation. It outperforms general active learning methods and achieves comparable performance to state-of-the-art E2E-AD methods using only 30% of the nuScenes dataset. The method includes an initial sample selection based on Ego-Diversity, which considers factors like weather, lighting, and vehicle speed. It also introduces three criteria for incremental selection: Displacement Error, Soft Collision, and Agent Uncertainty. These metrics help identify challenging and critical data samples for planning. Extensive experiments show that ActiveAD significantly improves planning performance and reduces annotation costs. The method is effective in various scenarios, including rainy or nighttime conditions and overtaking maneuvers. ActiveAD demonstrates strong robustness and excels in challenging situations, achieving comparable performance with 30% of the data compared to using the entire dataset. The method provides a data-centric perspective for E2E-AD and has the potential to inspire future research in this area.ActiveAD is a planning-oriented active learning method designed to improve sample and label efficiency in end-to-end autonomous driving (E2E-AD). The method addresses the challenge of high-quality labeled data scarcity and the long-tailed distribution of data in AD tasks. ActiveAD introduces novel diversity and uncertainty metrics based on planning-oriented philosophy to select the most useful samples for annotation. It outperforms general active learning methods and achieves comparable performance to state-of-the-art E2E-AD methods using only 30% of the nuScenes dataset. The method includes an initial sample selection based on Ego-Diversity, which considers factors like weather, lighting, and vehicle speed. It also introduces three criteria for incremental selection: Displacement Error, Soft Collision, and Agent Uncertainty. These metrics help identify challenging and critical data samples for planning. Extensive experiments show that ActiveAD significantly improves planning performance and reduces annotation costs. The method is effective in various scenarios, including rainy or nighttime conditions and overtaking maneuvers. ActiveAD demonstrates strong robustness and excels in challenging situations, achieving comparable performance with 30% of the data compared to using the entire dataset. The method provides a data-centric perspective for E2E-AD and has the potential to inspire future research in this area.