Abstract: Direct generalization from urban to rural settings is not a safe strategy for autonomous driving to handle unique challenges of rural driving scenarios. For example, passing through an intersection in rural environments may need to handle less attentive drivers with potentially higher approaching speeds, limited lighting, and obstruction of line of sight due to natural landscapes as well as differences in intersection geometry. This indicates that accurately detecting and predicting the intentions of dynamic traffic participants is crucial for ensuring safety and making collision-free decisions at rural intersections. In this work, we propose the development of learning-based methods for autonomous vehicles that can manage uncertainties and generate safe trajectories with the focus on rural intersections. Deep reinforcement learning methods will be considered for behavior planning for autonomous vehicles in the presence of dynamic participants with unpredictable intentions. The proposed method will be evaluated using the CARLA simulation environment. The developed model will take observation states from the simulation environment and sends control actions to the autonomous vehicle.