Individuals who suffer from paralysis as a result of a spinal cord injury list restoration of arm and hand function as a top priority. FES helps restore movement using the user's own muscles, but does not produce accurate and repeatable movements necessary for many functional tasks. Robots can assist users in achieving accurate and repeatable movements, but often require bulky hardware to generate the necessary torques. We propose sharing torque requirements between a robot and FES to reduce robot torque output compared to a robot acting alone, yet maintain high accuracy. Cooperative PD and model predictive control algorithms were designed to share the control between these two torque sources. Corresponding PD and MPC algorithms that do not use FES were also designed. The control algorithms were tested with 10 able-bodied subjects. Torque and position tracking accuracy were compared when the system was commanded to follow a functional elbow flexion/extension trajectory. The robot torque required to achieve these movements was reduced for the shared control cases compared to the algorithms acting without FES. We observed a reduction in position accuracy with the MPC shared controller compared to the PD shared controller, while the MPC shared controller resulted in greater reductions in torque requirements. Both of these shared algorithms showed improvements over existing options, and can be used on any given trajectory, allowing for better transferability to functional tasks.