<p>In the context of autonomous driving systems at mining sites, the planning of parking trajectories is crucial for ensuring the efficiency of loading tasks. Most studies predominantly employ the shortest path rule to determine the resulting trajectory. Different mining areas may have distinct requirements such as control maneuverability, travel distance, and interaction with other trucks. Existing planners have notable limitations as they require considerable manual effort to design an evaluation function that selects the optimal parking trajectory considering multiple factors. To address the limitations, this study proposes a novel approach for generating human-like, personalized parking trajectories. Specifically, the human-like planner transforms the continuous parking behavior observed in human drivers into a discrete set of maneuvers, deriving an internal evaluation function for parking behavior based on real-world human driving data. Building on this, an efficient parking trajectory is developed using bi-directional heuristic search to address queued loading in narrow areas. Extensive experiments conducted in real mining site environments demonstrate that the human-like parking trajectory planner closely mimics human drivers, reducing waiting time by 30% compared to traditional methods. Moreover, the proposed planner has been successfully applied in practical autonomous driving operations at mining sites. The source code implementation will be released as open-source: <a href="https://github.com/AMTbuaa/Human-like-parking-trajectory-planning">https://github.com/AMTbuaa/Human-like-parking-trajectory-planning</a>.</p>