Why are the force capabilities and controllability different for the End-Point and Exoskeleton robots?
The reasons for the different force capabilities between the two robots are two-fold. The first difference is due to the internal mechanics of the two robots such that the End-Point can produce 50% more torque. The second difference is due to the length of the linkage segments and equation relating forces and torques ( F = T / r ). With the KINARM End-Point robot, the linkage lengths are fixed, so the forces at the handle are the same for any subject. However, with the KINARM Exoskeleton, the linkage lengths are adjusted for subject size. As a subject gets taller (i.e. with longer arm lengths), then the force that the Exoskeleton can produce at the hand (or fingertip) gets smaller because the magnitude of possible torque stays constant.
The reason for the differences in controllability are completely independent from the issues of force production, and are entirely due to the differences in compliance between the End-Point and Exoskeleton robots. The KINARM End-Point robot is a much stiffer robot, which enables much higher gains to be implemented before the system becomes unstable. Because the limitation of the Exoskeleton is in the compliance of the mechanical design itself and not the controller, the only way around this problem is to either have a completely different robotic design (e.g. the End-Point robot) or to add other custom feedback devices and use those. We have not done this latter suggestion ourselves, but at least one of our customers has added an accelerometer to the Exoskeleton linkage to improve the kinematic feedback to the controller (they were attempting inertial compensation and were able to get ~50% compensation). The
theory behind this approach is that by placing an accelerometer out on the linkage, that feedback then bypasses most of the robotic compliance, thereby allowing higher gains in the feedback control. This approach is not trivial, unless you have extensive engineering experience, and although it would improve the controllability, we cannot provide any guidance on what sort of gains you might derive from it for different feedback/control types.