This paper presents a hybrid safe learning-based path planning algorithm that integrates classical graph-based planning with deep reinforcement learning techniques. The algorithm dynamically adjusts its strategy based on episode count, leveraging the strengths of both approaches to improve adaptability and safety in autonomous robotic navigation. Safety parameters, including minimum and average distances from obstacles, were consistently enhanced with the hybrid algorithm. Additionally, the results demonstrated reduced runtimes, emphasizing computational efficiency. Visual results further supported the algorithm’s effectiveness. Overall, the proposed hybrid algorithm offers a comprehensive solution to the challenges of path planning for autonomous robotic vehicles, contributing to advancements in safety, efficiency, and adaptability.