A loss of a limb can create significant difficulties in the everyday tasks of a person, being worse in more uncommon situations such as running or jumping. This creates problems with people that need to undergo an amputation, taking a toll on their wellbeing. With the tens of millions of amputees in the world and their number increasing every year, the development of prosthetic devices that are able to emulate a human limb becomes a necessity. The development of prostheses has improved over the years, by a considerable amount, however, most prostheses commercially available nowadays, are passive and therefore are not able to provide the necessary amount of energy to recreate proper human gait. To overcome this problem, active powered prostheses are being researched across the world. This thesis focuses on the development of a powered transtibial prosthesis’s controller based on the concept of impedance control. This type of controller tries to recreate the type of control found in humans, where a limb’s output impedance can be altered to allow for a more adaptable motion. The prosthesis was designed to work as a standalone device and use its sensors and a finite state machine to alter the output impedance of the device. The prosthesis’ prototype was designed to be worn by the Darwin-OP humanoid robot, which has a walking gait similar to that of a human. The controller requires the optimization of several parameters. For this task, a genetic algorithm was employed to help determine the controller parameters, as well as the parameters for the Darwin-OP waking motion controller, this one being based on Central Pattern Generators. A comparison between adding or not, a dynamic model of the actuator’s body to the controller, was performed. This comparison provided information about the advantages and disadvantages of using a model on the controller and led to the choice of foregoing the model in the transtibial prosthesis. In the end, the optimization was able to determine the parameters for the prosthesis controller and the simulated Darwin-OP robot, with the device. The humanoid was able to walk forward, albeit at a slower speed than the humanoid without the prosthesis. In regard to the controller itself, it was able to accurately detect the stance and swing phases of the device. Tests of the controller were performed in simulation using Webots™.