Soft-Robotics is a new field research field, born to overcome the limitations of traditional rigid robots, soft-robots in particular can ensure a safer machine-human interaction and have the flexibility to maneuver in unpredictable and challenging circumstances. BeCause of their structure and the materials used in their construction, they are characterized by nonlinear behavior, thus resulting hyperedundant and underactuated. This suggests that the relationship between the task space and the actuations space is challenging to define and difficult to govern. The objective of this work was to employ artificial intelligence to train a soft robot to throw objects in the direction of a predetermined target, through the completion of a circular trajectory by the robot, in order to exploit the moment of inertia..