This paper reviews recent advances in Rapidly-exploring Random Tree (RRT) algorithms from 2021 to 2023, focusing on theoretical improvements and practical applications. RRT is a path planning algorithm that combines search and sampling properties, making it effective for generating high-quality paths that balance global and local optima. Theoretical improvements include branching strategy, sampling strategy, post-processing, and model-driven RRT. Applications span welding robots, assembly robots, search and rescue robots, surgical robots, free-floating space robots, and inspection robots. Challenges include designing hyperparameters, weak generalization, and hardware reliability in large-scale unstructured environments. Future trends include multi-robot collaboration, human-robot collaboration, real-time path planning, self-tuning hyperparameters, and hardware design. RRT-based algorithms face challenges in real-time performance, uncertain environments, and path quality. Improvements include branching strategies like RRT*, RRT-connect, and B-RRT*, sampling strategies like informed-RRT* and Quick-RRT*, and post-processing techniques like path smoothing and curvature-aware planning. Model-driven RRT incorporates machine learning for path optimization. Despite these advancements, RRT still struggles with scalability, real-time performance, and generalization in complex environments. Future research should focus on improving RRT's adaptability, robustness, and integration with other technologies like AI and robotics.This paper reviews recent advances in Rapidly-exploring Random Tree (RRT) algorithms from 2021 to 2023, focusing on theoretical improvements and practical applications. RRT is a path planning algorithm that combines search and sampling properties, making it effective for generating high-quality paths that balance global and local optima. Theoretical improvements include branching strategy, sampling strategy, post-processing, and model-driven RRT. Applications span welding robots, assembly robots, search and rescue robots, surgical robots, free-floating space robots, and inspection robots. Challenges include designing hyperparameters, weak generalization, and hardware reliability in large-scale unstructured environments. Future trends include multi-robot collaboration, human-robot collaboration, real-time path planning, self-tuning hyperparameters, and hardware design. RRT-based algorithms face challenges in real-time performance, uncertain environments, and path quality. Improvements include branching strategies like RRT*, RRT-connect, and B-RRT*, sampling strategies like informed-RRT* and Quick-RRT*, and post-processing techniques like path smoothing and curvature-aware planning. Model-driven RRT incorporates machine learning for path optimization. Despite these advancements, RRT still struggles with scalability, real-time performance, and generalization in complex environments. Future research should focus on improving RRT's adaptability, robustness, and integration with other technologies like AI and robotics.