This research presents a novel approach for geometrically constrained path planning. The methodology introduced is based on the standard Fast Marching Square (FM2) method and a path extraction approach based on an optimisation process named Differential Evolution (DE). The geometric constraints are introduced in the path extraction phase. This step uses both the funnel potential of the environment created with FM 2 and the geometric constraints as a cost function to be minimised. The use of an optimisation process permits to get a close-to-optimal path, while mostly keeping the characteristics of the paths computed with FM 2 . In the presented simulations, two kinds of restrictions have been applied: soft and hard ones. The first allows some flexibility in the constraints, while the last force the constraints to be met along all the path. The method has been tried with a simple bar, an articulated double bar and a finger-like kinematic chain in different environments. Results show the potential of this method in constrained path computation.