The focus of this work is the real-time prediction and compensation of static deflections in robotic manipulator arms. A general manipulator deflection model is developed based on static beam theory and robot kinematics. An optimization technique is proposed to determine the stiffness of the manipulator components using end-effector deflection measurements. Strategies for incorporating this modeling approach into a manipulator controller are also presented along with the results of a successful application of this research. This work is an extension of previous manipulator deflection research. Multiple pairs of torsional stiffness elements and beam elements are used to model complex link and joint geometries whereas previous models only used a single beam per manipulator link. In addition, the modeling algorithms and stiffness characterization methods are general and may be applied directly to any serial manipulator. Also, the optimization techniques used to characterize a manipulator's stiffness provide a more accurate and flexible approach than previous analytical methods. The deflection model was successfully tested using a nuclear steam generator service manipulator. Since this manipulator is considerably more flexible than common industrial robots, it serves as a near worst-case test for deflection modeling. The end effector was found to deflect as much as 1.5 inches due to the weight of the links and joints. The deflection model was able to compensate for 94% of the end-effector deflection, allowing the manipulator to perform tasks requiring a positioning accuracy of 0.09 inches. The algorithms for flexible forward and inverse kinematics as well as trajectory generation were incorporated directly into the manipulator controller code. These modules were capable of running in real-time with little computational expense. / Master of Science