I am stuck with an optimisation problem. In brief; I am trying to optimise the weights associated to the control points of a NURBS 3D curve using either a genetic algorithm or a PSO. The cost function is based on some path preferential characteristics such as length , curvature constraint,feasibility (in terms of obstacle avoidance etc)

I feel the results of the optimisation with my current attempts (using the Optimisation Toolbox of Matlab) strongly depend on the parameter settings of the algorithm I am using. Ideally I would like to define my cost function in such a way that the optimisation tends always to the global minimum within some constraints (like curvature and obstacle avoidance).

What I am trying to do is quite similar to what they did in this paper (https://link.springer.com/content/pdf/10.1007/978-3-319-19264-2_19.pdf ) but it is not clear how they can guarantee the constraints.

does anybody know how to solve this?

Thank you for your attention,