• gait generation
• iterative learning control
• motion phases
• optimal control
• planning with contacts
• switched dynamics
• trajectory optimization

The problem is faced through a trajectory optimization that takes into account the dynamics of the
system, split in dierent motion phases. The system is represented through a 2D centroidal dynamics
model. Feet positions are included into the formulation. Terrain and friction cone constraints are
enforced to guarantee a physically feasible motion. Then, the obtained base and feet positions are
mapped into joints positions and velocities through inverse kinematics. To make the simulation
and the real robot run, a ROS2 hardware interface is used. To ll the gap between simulation and
validation experiments in order to achieve the desired task, Iterative Learning Control is exploited. In
order to let the real robot learn the task. Thanks to this implementation the potential of the proposed
algorithm is tested both in simulation and experimentally on the quadrupedal robot Solo-12.