Manipulability analysis of humanoid robots with redundant arms is difficult due to the presence of large number of degrees-of-freedom (dof). Most researchers address manipulability issues without considering the effects of joint limits, obstacles and singular spaces in a Cartesian workspace. Hence, development of an accurate manipulability analysis technique, which can increase task performance by considering the above-mentioned issues is crucial for completing cooperative and non-cooperative tasks. Our paper proposes a new approach for determining manipulability measurements of a humanoid robot with redundant arms doing coordinated and non-coordinated tasks by analysing manipulability ellipsoids constructed through a desired trajectory. Penalty functions for compensating joint limits and avoiding obstacle regions are multiplied along with a Jacobian matrix to generate an Augmented Jacobian matrix. Manipulability ellipsoids determined using the Augmented Jacobian for individual configurations are compared with desired manipulability ellipsoids for finalizing the joint solutions. The advantages of proposed approach over conventional approach and significance of employing proposed approach for updating joint configurations are presented in this paper. The experimental validation of the proposed method using a developed humanoid robot is also given in this paper.