This paper considers the generation of trajectories to a given protein conformation and presents a novel approach based on artificial potential functions--originally proposed for multi-robot navigation. The artificial potential function corresponds to a simplified energy model, but with the novelty that--motivated by work on robotic navigation--a nonlinear compositional scheme of constructing the energy model is adapted instead of an additive formulation. The artificial potential naturally gives rise to a dynamic system for the protein structure that ensures collision-free motion to an equilibrium point. In cases where the equilibrium point is the native conformation, the motion trajectory corresponds to the folding pathway. This framework is used to investigate folding in a variety of protein structures, and the results are compared with those of other approaches including experimental studies.