Human Centered Robotics
Our semester long project focused on replicating simulated acrobatic maneuvers using Boston Dynamics' Atlas. Using Centroidal Dynamics, a simplified model of the robot's mass distribution was used to calculate the linear and angular trajectory of the robot. Model Predictive Control was then used to control the non-linear dynamics of the robot as it attempted to follow the calculated desired trajectory.
This is an implementation of Atlas utilizing Whole-Body Control (WBC) and Divergent Component of Motion (DCM) to balance while walking in a straight line. The DCM planner is given a set of foot positions to achieve a walking motion and gait timing to ensure a human-like gait. The feet reaction forces and joint accelerations are minimized to reduce high-energy movements.