In this paper, we develop a human-robot cooperative control system to aid the independence of quadriplegic and other mobility impaired individuals. The hardware of the system consists of a 14-DOF dual-arm robot mounted on a wheelchair. The full system is commanded by the user using a sip-puff device to control any 3-DOF motion of the mobile manipulator system. A novel human-robot cooperative control strategy is proposed to combine the 3-DOF user input with autonomous control for this highly redundant system. The human input commands are formulated as soft equality constraints, while the obstacle avoidance, singularity avoidance and workspace limitation are formulated as inequality constraints based on the potential field functions. A new redundancy resolution control scheme is designed such that when the system is far from the inequality constraint boundaries, the human input will be followed. Near the constraint boundary, the human input commands will be automatically modified to prevent constraint violation. We show that such redundancy resolution scheme guarantees that the inequality constraints are always satisfied. Experimental results are included to demonstrate the effectiveness of the proposed scheme.