A Non-Smooth Numerical Optimization Approach to the Three-Point Dubins Problem (3PDP)

Abstract

This paper introduces a novel non-smooth numerical optimization approach for solving the Three-Point Dubins Problem (3PDP). The 3PDP requires determining the shortest path of bounded curvature that connects given initial and final positions and orientations while traversing a specified waypoint. The inherent discontinuity of this problem precludes the use of conventional optimization algorithms. We propose two innovative methods specifically designed to address this challenge. These methods not only effectively solve the 3PDP but also offer significant computational efficiency improvements over existing state-of-the-art techniques. Our contributions include the formulation of these new algorithms, a detailed analysis of their theoretical foundations, and their implementation. Additionally, we provide a thorough comparison with current leading approaches, demonstrating the superior performance of our methods in terms of accuracy and computational speed. This work advances the field of path planning in robotics, providing practical solutions for applications requiring efficient and precise motion planning.