The objective of this thesis is to show that warping can be used in a thin-walled beam to create a suitable replacement for a classical differential mechanism. This normally negated deformation of the cross section can be used to create an inverse transmission mechanism, used to create a monolithic and simple alternative to a classic differential mechanism. This warping beam differential mechanism can be used to add walking functionality to a back support passive exoskeleton. These exoskeletons are optimized to reduce muscle stress during lifting, but this comes with the downside of an increase in muscle activity during walking. To counter this increase in muscle activity, the use of a differential mechanism is suggested. To show the feasibility of the proposed solution, a characterization of the warping beam differential mechanism is made for different geometric properties. This characterization is done with respect to two functionalities, to determine the geometric advantage of the inverse transmission and the rotational compliance under torque of the beam. The warping constant and torsion constant are determined to be the geometric properties of most influence on the behaviour of the warping beam. It is also shown that within the proposed boundary conditions, different beam cross sections with the same geometric properties show the same behaviour. An increase in warping constant results in an increase in geometric advantage and a reduction of the rotational compliance. The torsion constant has a linear correlation to the stored energy for the inverse transmission functionality, having less influence on the geometric advantage and rotational compliance compared to the warping constant. This thesis shows that this new kind of differential mechanism can be used to replace the classical mechanism and the potential can be seen with an achieved theoretical geometric advantage of almost 1. A systematic overestimation of the geometric advantage and an underestimation of the compliance is made by the model with respect to the experiments, but with some improvements this theoretical goal is not far away.