• iterative learning control
• planning optimization
• quadruepd locomotion

Legged robots are capable of crossing challenging and uneven terrain, allowing the development of new applications, like robotic environmental monitoring. This notwithstanding, the locomotion of these robots is often limited to less dynamic walking and running gaits. Instead, jumping is a highly dynamic motion that can theoretically allow a legged robot to conquer obstacles that could not be easily reached by the standard aforementioned gaits.
In this context, this thesis proposes a locomotion planning and control framework for achieving forward jumping with quadrupedal robots.
The motion planner uses direct trajectory optimization on the quadruped sagittal plane and a single rigid body approximation of the dynamics to generate a feasible motion of the robot to reach the desired position over an obstacle while avoiding collision with it. Then, the obtained base trajectory is mapped into joint position, velocity, and torque references, which can be executed on the real system by any joint controller. Given the planner output, Iterative Learning Control is used to fill the gap between the approximated model employed in the planner and the actual physical system.
To validate the proposed method, the quadrupedal robot SOLO12 has been chosen as the testing platform. To this end, a ROS2 hardware interface has been developed to make available the robot’s hardware features in the framework, making possible the execution of the control loop on the real system. Thanks to the implementation of this ROS2 interface, the performance of the proposed method is validated both in simulation and on the real robot. Experimental results show that the framework enables the robot to jump over obstacles up to a height of 15 cm.