Motivated by the redundant actuation mode of the human body, which produces several unique functions that general robotic mechanisms cannot generate, an anthropomorphic robot module resembling the musculoskeletal structure of the human upper limb is investigated. Specifically, optimal kinematic design of the proposed robot module is treated. Initially, the kinematic model of the robot module is obtained, and five design indices are then defined based on the kinematic model. These indices include workspace area, Jacobian isotropic index, maximum force transmission ratio, and gradient design indices for the Jacobian isotropic index and the maximum force transmission ratio. To deal with this multi-criteria based design, a kinematic composite design index (KCDI) is introduced which combines several individual design indices as a unique design index using the max-min principle of fuzzy theory. Two KCDIs are considered as objective functions in the kinematic optimization. One includes the three design indices without the two gradient design indices, and the other includes all five design indices. As a result of optimization based on the two KCDIs, two sets of nine optimal kinematic design parameters are obtained. The two designs are compared with respect to three operational performances: maximum load handling capacity, maximum hand velocity and maximum hand acceleration capability. Optimal actuator sizes of the two designs are also compared for the given specifications of three operational performances. The design taking into account all five design indices shows superior operational performances and smaller actuator sizes than those of the design ignoring gradient design indices. (C) 1997 Elsevier Science Ltd.