Sample-Based Model Predictive Control for Stewart Platform using Data-Driven Model

Stewart platforms stand out by offering six degrees of freedom, surpassing traditional gimbal systems that are limited to rotational corrections such as roll, pitch, and yaw. Gimbal systems, typically integrated with cameras and sensors, are used for recording footage in aircraft, helicopters, and UAVs (unmanned aerial vehicles) and as handheld devices for photography. Their applications include image tracking, surveillance, target tracking, and engagement, particularly in industrial and military contexts. The addition of three degrees of freedom for corrective movements allows the Stewart platform to significantly enhance existing motion correction systems. However, the Stewart platform's full potential is difficult to achieve due to the absence of forward kinematics information, unlike gimbals, which are series robots. To address this limitation, we propose a data-driven control framework using sample-based model predictive control. However, similar to other parallel robots, the forward kinematics of the Stewart platform is complex because of the constraints at the joints. Additionally, each actuator of the Stewart platform can easily reach its saturation limit in addition to the uncertain dynamics. Therefore, we propose using sample-based model predictive control to optimize control over a finite time horizon while avoiding the constraint of maintaining the linear actuators by utilizing sample trajectories.