High-definition (HD) Map systems will play a pivotal role in advancing autonomous driving to a higher level, thanks to the significant improvement over traditional two-dimensional (2D) maps. Creating an HD Map requires a huge amount of on-road and off-road data. Typically, these raw datasets are collected and uploaded to cloud-based HD map service providers through vehicular networks. Nevertheless, there are challenges in transmitting the raw data over vehicular wireless channels due to the dynamic topology. As the number of vehicles increases, there is a detrimental impact on service quality, which acts as a barrier to a real-time HD Map system for collaborative driving in Autonomous Vehicles (AV). In this paper, to overcome network congestion, a Q-learning coverage-time-awareness algorithm is presented to optimize the quality of service for vehicular networks and HD map updates. The algorithm is evaluated in an environment that imitates a dynamic scenario where vehicles enter and leave. Results showed an improvement in latency for HD map data of $75\%$, $73\%$, and $10\%$ compared with IEEE802.11p without Quality of Service (QoS), IEEE802.11 with QoS, and IEEE802.11p with new access category (AC) for HD map, respectively.