Control of an Emika Franka Panda robotic arm

The handling of suspended loads is a major challenge in industrial automation. Rapid transportation often induces oscillations in underactuated cranes, degrading accuracy and safety. While load-swing suppression is widely studied, experimental validation is typically performed on custom-built setups, preventing direct comparison between control strategies. To address this, the primary aim of this thesis is to develop a general prototype for testing different control laws for rotary cranes. This is achieved through two main contributions. First, a versatile experimental setup is developed as a standardized benchmark platform. The system features a custom two-degree-of-freedom spatial pendulum attached to a 7-DoF Franka Emika Panda manipulator, operating within a 1 kHz hard real-time framework based on a real-time Linux kernel and ROS. Second, using this prototype, a novel anti-sway cascade control strategy is proposed and compared with a baseline Partial Feedback Linearization (PFL) approach. The joint-space PFL guarantees stability under nominal conditions but deteriorates during dynamic maneuvers, where transient accelerations cause significant load oscillations. Moreover, formulating PFL for redundant robotic arms is mathematically challenging. To overcome these limitations, the proposed cascade architecture operates in the Cartesian space. An outer loop anticipates base acceleration and computes the lean angles required to balance payload dynamics, while an inner LQR controller generates accelerations to maintain this attitude. Commands are converted into joint torques via Operational Space Control. Experimental results validate the prototype and demonstrate the superiority of the cascade approach, maintaining payload angular deviations below 2.5° during complex trajectories and exhibiting robust recovery from disturbances.