Path planning is a key and complex element for every unmanned ground vehicle development. Once the 3D reconstruction of the environment is completed and the objective configuration (desired position and pose) is defined, there has to be a careful path planning algorithm. That path is subject to many restrictions: it has to be time optimal; we have limited degrees of freedom to work with since the vehicle is a non-holonomic robot; we have limited computational power and real-time constraints regarding on-board equipments; and finally the vehicle's mechanical limitations, like the maximum curvature.In this paper we present a new methodology for the path planning calculation. It was meant to be a one for all methodology, useful for different scenarios (automotive, industrial applications, mining, etc.) and different platforms (car-like vehicles, forklift trucks, etc.).This paper splits the problem in two stages. The first one faces the problem of reaching the goal with an a priori knowledge of the position affected by noise. The second approach develops a system capable of reaching the goal, enhancing the precision using a detection system, mainly based on computer vision. Particular focus is given to the interaction between the two methods proposed.

1001

996

6