Accurate alignment of knee implants is essential for the success of total knee arthroplasty. Although mechanical alignment guides have been designed to improve alignment accuracy, there are a number of fundamental limitations of this technology that will inhibit further improvements. Computer-assisted surgical navigation systems have been developed to provide surgeons the ability to make accurate, reproducible installation of implants and record quantitative information such as range of motion and joint laxity. We worked with the Medical Robotics and Computer-Assisted Surgery Laboratory from the Robotics Institute at Carnegie Mellon University and the Institute for Computer-Assisted Orthopaedic Surgery in Pittsburgh, Pennsylvania, to develop a navigation system for total knee arthroplasty. Our research team at Stanford designed a "quick disconnect" coupler to provide a reliable means of physically attaching and detaching reference frames to bone, a step critical to the success of surgical navigation. We have evaluated the performance of techniques that estimate the location of the center of the hip and ankle joints, key landmarks used by navigation systems. We developed and evaluated a "pivoting" algorithm that accurately estimates the location of the hip joint center and is minimally affected by noisy motion data. We also tested and refined methods that locate the center of the ankle. These studies improved the core algorithms for navigation.