This paper proposes a Priority-Aware Collaborative Perception (PACP) framework for connected and autonomous vehicles (CAVs) to enhance perception accuracy and communication efficiency. The framework addresses the limitations of traditional collaborative perception methods, which often prioritize fairness in data transmission but neglect the varying importance of individual vehicles. PACP introduces a BEV-match mechanism to determine the priority levels of CAVs based on their correlation with the ego vehicle and the quality of their perception data. By leveraging submodular optimization, PACP finds near-optimal transmission rates, link connectivity, and compression metrics. Additionally, a deep learning-based adaptive autoencoder is deployed to modulate image reconstruction quality under dynamic channel conditions. The framework also incorporates a priority-aware mechanism that dynamically adjusts the importance of data based on real-time channel quality and analytics, reducing unnecessary data processing and enhancing raw-level sensing data fusion and system responsiveness. Experimental results demonstrate that PACP outperforms existing methods by 8.27% and 13.60% in terms of utility and precision of the Intersection over Union (IoU). The proposed framework is evaluated on a CAV simulation platform, CARLA, with the OPV2V dataset, showing significant improvements in performance. The key contributions of PACP include the first implementation of a priority-aware collaborative perception framework, the application of submodular theory in a two-stage optimization framework, and the integration of a deep learning-based adaptive autoencoder. The framework addresses the challenges of data-intensive transmissions under dynamic and constrained channel capacities, ensuring efficient communication and perception in autonomous driving scenarios.This paper proposes a Priority-Aware Collaborative Perception (PACP) framework for connected and autonomous vehicles (CAVs) to enhance perception accuracy and communication efficiency. The framework addresses the limitations of traditional collaborative perception methods, which often prioritize fairness in data transmission but neglect the varying importance of individual vehicles. PACP introduces a BEV-match mechanism to determine the priority levels of CAVs based on their correlation with the ego vehicle and the quality of their perception data. By leveraging submodular optimization, PACP finds near-optimal transmission rates, link connectivity, and compression metrics. Additionally, a deep learning-based adaptive autoencoder is deployed to modulate image reconstruction quality under dynamic channel conditions. The framework also incorporates a priority-aware mechanism that dynamically adjusts the importance of data based on real-time channel quality and analytics, reducing unnecessary data processing and enhancing raw-level sensing data fusion and system responsiveness. Experimental results demonstrate that PACP outperforms existing methods by 8.27% and 13.60% in terms of utility and precision of the Intersection over Union (IoU). The proposed framework is evaluated on a CAV simulation platform, CARLA, with the OPV2V dataset, showing significant improvements in performance. The key contributions of PACP include the first implementation of a priority-aware collaborative perception framework, the application of submodular theory in a two-stage optimization framework, and the integration of a deep learning-based adaptive autoencoder. The framework addresses the challenges of data-intensive transmissions under dynamic and constrained channel capacities, ensuring efficient communication and perception in autonomous driving scenarios.