• quadruped robot
• optimal control
• ROS
• trajectory optimization
• motion planning
• legged robot
• ROS2

Legged robots are attracting more and more interest due to their wide mobility capabilities over any kind of terrain or environment; this has led the scientific community to investigate new planning and control techniques or adapt already-known ones. On the other hand, the high complexity of these systems and the inability to directly generate motion, except through interaction with
the outside world, pose a great challenge when dealing with such robots. Building on
the theory of Trajectory Optimization, this thesis proposes a new approach to
simultaneously plan the base trajectory and contact sequence in an optimal manner,
applying it to the case of the longitudinal motion of a quadruped robot. The resulting trajectory is next translated into joint position, velocity and torque references, which can be used by any joint controller to operate the real system. Iterative Learning Control is then used to close the gap between the approximated model used in the planner and the real physical system, according to the planner output. In the end, the proposed method was tested on the quadruped robot SOLO12 using a custom ROS2 interface, which allowed to validated its performances both in simulation and on the real robot. Experiments show that the method enables the robot to perform complex motions, like galloping at different speeds.