This paper proposes a technique for animating simulated characters to perform controlled
steps. The desired step is controlled by high-level goals, namely step position and step
duration. These stepping goals guide the desired time-varying values for the center of mass
and the stepping foot which in turn lead to objectives dictating the desired changes in
momentum and joint angles over the duration of the step. Our approach employs a
multiobjective optimization to solve for joint accelerations from the objectives and uses …