Adding the Feedback Loop

Feedforward-only control will fail as soon as the commanded state no longer matches the operational state of the real system. With inaccurate assessment of the operational state, all estimates of actuation torques/forces will be incorrect. In the case of our robot, the state is defined by the position and velocity for each of our actuation axes -- you might say the position and speed for each motor in our robot. For an unstable mechanical system like our robot, it has been shown in our previous videos that it is impossible to anticipate motion response to given inputs beyond 1.5 seconds. Of course, this is well understood in control theory, and the strategy is to simply remain well informed of the state of the system, which is achieved through the feedback from the sensors. This video demonstrates how to employ feedback, and shows how beautifully this works to enable control of the complex mechanics of our robot.