Powered prosthetic knees offer many improvements over passive devices; however, the added actuation is difficult to control due to the lack of input from the user. A means of controlling a powered prosthetic knee is proposed by predicting the type of swing behavior the knee must perform based on the position of the foot relative to the person during toe-off. The software to accomplish this is implemented via a finite state machine which activates specific knee angle reference generators for each state. The reference generators serve as the input of a nonlinear feedback controller to ensure accurate positioning of the knee joint.