Dynamic Curvature Constrained Path Planning

TL;DR

The paper tackles 2D path planning under curvature constraints in constrained environments by introducing the Dynamic Curvature-Constrained Path Planning Algorithm (DCCPPA). It combines a node-based graph with obstacle-perimeter sampling and a hybrid global-local search that incorporates directional curvature via the objective $J(p)=\text{Path Length}(p)+\beta\cdot\text{Curvature Deviation}(p)$ and curvature constraints $\kappa\le\text{Curvature Threshold}$, delivering curvature-aware trajectories while ensuring collision avoidance. The authors benchmark DCCPPA against PRM and RRT, demonstrating reduced path-step counts (e.g., averages around 48 for DCCPPA vs ~53 for RRT and ~284 for PRM across 10 trials), and provide analyses of time $O(N\cdot M)$ and space $O(N+M)$ complexities. The work suggests that DCCPPA offers improved efficiency and robustness in 2D planning with curvature constraints and highlights its potential for extension to 3D, dynamic, and multi-agent settings, with opportunities for machine learning enhancements and better visualization tools.

Abstract

Effective path planning is a pivotal challenge across various domains, from robotics to logistics and beyond. This research is centred on the development and evaluation of the Dynamic Curvature-Constrained Path Planning Algorithm (DCCPPA) within two dimensional space. DCCPPA is designed to navigate constrained environments, optimising path solutions while accommodating curvature constraints.The study goes beyond algorithm development and conducts a comparative analysis with two established path planning methodologies: Rapidly Exploring Random Trees (RRT) and Probabilistic Roadmaps (PRM). These comparisons provide insights into the performance and adaptability of path planning algorithms across a range of applications.This research underscores the versatility of DCCPPA as a path planning algorithm tailored for 2D space, demonstrating its potential for addressing real-world path planning challenges across various domains. Index Terms Path Planning, PRM, RRT, Optimal Path, 2D Path Planning.

Figures (3)

• Figure 1: Flow Chart of Research Methodology
• Figure 2: Obstacles with start and goal point for scenario 1
• Figure 3: Path construction of DCCPPA algorithm