<p>Global climate change is amplifying the uncertainty of marine environments, rendering traditional path planning algorithms, which rely on static heuristics, increasingly inadequate and often perilous for maritime navigation. Their inability to adapt to real-time, synergistic risks like typhoons and abnormal currents represents a critical gap, leading to suboptimal routes that compromise both safety and efficiency. To bridge this gap, this study introduces QLA*, a novel algorithm that pioneers a learning heuristic function by deeply integrating Q-learning with the A* search. This core innovation transforms the heuristic from a static guide into a dynamic, experience-driven mechanism, enabling the planner to emulate an expert mariner’s intuition by dynamically adjusting its strategy based on evolving sea conditions. QLA* is further enhanced by an intelligent search-learning loop and a dual-trigger replanning strategy that leverages learned Q-values to drastically reduce computational redundancy. Extensive experiments demonstrate QLA*’s superiority over established algorithms. Crucially, a case study on the "Milano" vessel accident confirms its real-world efficacy, showing that the proposed algorithm could have successfully navigated the vessel away from the high-risk areas that led to the disaster. By creating a robust framework that intelligently balances safety and efficiency, QLA* offers a significant contribution towards the autonomous and intelligent transformation of the shipping industry.</p>