The research objectives of this thesis are, to begin with, to investigate the geometrical significance of the characteristic length of serial robots used to calculate the condition number of their Jacobian matrix; then, to show how sequential quadratic programming (SQP) can be used as a redundancy-resolution algorithm for robots performing manufacturing tasks. As a performance objective the condition number of the normalized Jacobian matrix, also called the dexterity index, is used. This index is important, for it provides a measure of “distance” form singularities. By staying “away” from singularities, the robot should be able to perform tasks much more accurately. In fact, the condition number of the Jacobian bounds the ratio of a norm of joint errors to a norm of the end-effector non-dimensional pose errors. The dexterity index thus plays an important role in robot-assisted machining tasks, since accuracy is one of the main issues for these applications. One major concern in computing the condition number of the normalized Jacobian matrix lies in defining a characteristic length, needed for normalization purposes. Without this normalization, the condition number would not have any significance since, in its computation, quantities with units of length squared would be added to dimensionless quantities. To gain further understanding into the characteristic length and condition number, a geometric interpretation of these two items, the latter as pertaining to the characteristic length and condition number of the normalized Jacobian is investigated for three-degree-of-freedom (three-dof) planar robots. The procedure can be extended to six-dof spatial robots, but only for the interpretation of the condition number. In a second part, for manufacturing tasks such as machining, welding, deburring and milling, sequential quadratic programming (SQP) is applied as a redundancyresolution algorithm. For these tasks, an axis of symmetry exists on the end-effector, which is the tool axis. A rotation of the end-effector about this axis clearly does not influence the task at hand. When the task is executed by a 6-dof robot, the robot has an extra degree of freedom, allowing it to perform a secondary task. In this case, the robot is called functionally redundant. This redundancy arises only because of the task dimension, which is lower than the number of axes available. Concerning functionally redundant robots, special attention must be given to their redundancy-resolution, as conventional methods using the null space of the Jacobian matrix are not applicable. The SQP approach developed here takes this feature into account: it finds the Jacobian null space by identifying the tool axis in the robot base coordinates. This is an approach similar to the recently developed Twist Decomposition Algorithm (TWA); however, the SQP is expected to provide better convergence properties of the algorithm since a quadratic approximation of the objective function is used. A functionally redundant robot using the SQP redundancy-resolution method is also investigated in an example to show the effectiveness of the proposed SQP redundancy-resolution method. In this example, a comparison is made between the quasi-Newton and Newton-Raphson methods to find the posture of minimum condition-number for the robot. RobotMaster is used in this example to verify the trajectory.