Investigation of various path planning optimisation techniques from academic papers in Jupyter Notebook.
This notebook elaborates the testing and development of an offline path planning optimisation pipeline to generate a safe and feasible reference path for an ego vehicle.
The pipeline takes a set of coarsely placed waypoints and adjusts their localisation so as to comply with certain path curvature constraints in view of the vehicle’s size and steering capability. The adjusted waypoints should be as close as possible to the original waypoints without violating any curvature constraints.
A controlled real-world test was conducted only on the waypoints generated from the unconstrained approach. Empirical data indicated higher comfort levels during transportation when compared to the tracking of unoptimised waypoints. With curb information accounted for (see the controlled approach), the resultant path held a safer distance from the curb without any noticeable loss in passenger comfort.
pip install -r requirements.txt