Self-aligning mechanisms are essential components in facilitating adaptability in wearable robots, but their synthesis from scratch is very challenging. To overcome this hurdle, we propose a so-far-unprecedented autonomous method to synthesize self-aligning knee joint mechanisms, requiring neither a baseline design nor human intervention during synthesis. Our method transforms the synthesis problem into an optimization problem amenable to an efficient gradient-based algorithm using a discretized ground mechanism model. The main challenge in the conversion lies in how to define the objective and constraint functions in order to ensure the fundamental self-aligning capability and also to impose a desired force transmittance profile. Several design cases were considered to show the effectiveness of the newly proposed functions for the optimization-based synthesis formulation, notably in addressing degree-of-freedom requirements. Although this study focuses primarily on knee joint mechanisms assisting gait motion and aligning with the flexion axis, the developed method can be applied to other self-aligning robot mechanisms.