This article provides a comprehensive review of path planning for autonomous ground vehicles in unstructured environments, a critical area that lags behind structured environments due to the challenges posed by harsh conditions and complex vehicle-terrain interactions. The authors categorize path planning into hierarchical and end-to-end approaches, emphasizing the unique aspects such as terrain traversability analysis (TTA), cost estimation, and constraints. They discuss the factors influencing TTA, including terrain geometry, physical properties, and vehicle dynamics, and review various TTA methods. The article also covers cost estimation for safety, energy, and comfort, and constraints related to terrain and vehicle limitations. Global path planners, such as graph search, sampling-based, and artificial potential field methods, are discussed, along with local path planners that optimize paths based on global plans while considering vehicle and terrain constraints. The end-to-end approach, which directly outputs paths based on sensor data, is highlighted for its adaptability but noted for its reliance on large datasets and computational power. The article concludes by identifying key areas for future research, including handling multiple uncertainties in path planning and improving the adaptability of end-to-end methods.This article provides a comprehensive review of path planning for autonomous ground vehicles in unstructured environments, a critical area that lags behind structured environments due to the challenges posed by harsh conditions and complex vehicle-terrain interactions. The authors categorize path planning into hierarchical and end-to-end approaches, emphasizing the unique aspects such as terrain traversability analysis (TTA), cost estimation, and constraints. They discuss the factors influencing TTA, including terrain geometry, physical properties, and vehicle dynamics, and review various TTA methods. The article also covers cost estimation for safety, energy, and comfort, and constraints related to terrain and vehicle limitations. Global path planners, such as graph search, sampling-based, and artificial potential field methods, are discussed, along with local path planners that optimize paths based on global plans while considering vehicle and terrain constraints. The end-to-end approach, which directly outputs paths based on sensor data, is highlighted for its adaptability but noted for its reliance on large datasets and computational power. The article concludes by identifying key areas for future research, including handling multiple uncertainties in path planning and improving the adaptability of end-to-end methods.