Multi-modal assistive system is a hybrid system that aims to provide assistance-as-needed for paretic subjects. The hybrid system combines human muscle efforts, stimulated by a Functional Electric Stimulator (FES) and an lower-limb active exoskeleton. With optimal sharing of the two activation inputs, a hybrid system could overcome several challenges, such as human force due to FES-induced muscular fatigue. In this paper, a novel optimal controller, based on an efficient (3.9 ms to predict a time windows of 40 ms) closed-loop direct collocation approach is proposed to predict the control allocation between FES and exoskeleton. Simulation results, performed with dynamically varying desired trajectory, show that the proposed controller adapts the control input with respect to muscular fatigue.