Enhancing Sampling-Based Path Planning for Autonomous Robots in Confined Areas

This thesis investigates how different variants of the Rapidly-exploring Random Tree (RRT*) algorithm can be used to plan drivable paths for autonomous vehicles in confined environments. The focus is on comparing the standard version of RRT* with an improved version that uses informed sampling to make the search for the optimal path more efficient. Furthermore, the vehicle’s steering function is a major focus, the movement constraints of a real car, with the aim of generating physically drivable paths. The study was conducted in three different simulated environments: an alley, a warehouse space, and a U-turn, each representing different levels of complexity and constraints. The results show that the improved variant of RRT* led to noticeable improvements in some cases, especially in the open warehouse environment, where informed sampling effectively avoided unnecessary exploration. In simpler and more confined environments, such as the alley, the differences were small, as both algorithms produced similar solutions. Statistical analysis using the Fréchet distance and the number of nodes showed that informed sampling can reduce computational complexity and improve search outcomes in the right type of environment. However, the results are also influenced by the quality of the initial path — particularly in more complex scenarios like the U-turn. The conclusion is that informed RRT* has the potential to improve both the efficiency and robustness of path planning, but its benefits are highly dependent on the structure of the environment and the initial conditions.