Model Predictive Control-Based Trajectory Planning for a Stewart Platform: Planning and Tracking for UAV Capture Using Parallel Manipulator System
Authors
Posluszny, Michal Antoni ; Jensen, Rasmus ; Fogh, Anders Henriksen
Term
4. semester
Education
Publication year
2025
Submitted on
2025-06-02
Pages
104
Abstract
This project develops a trajectory planning and tracking system for a Stewart platform as part of an active UAV (drone) retrieval strategy to improve safety during autonomous landings. A Stewart platform is a six-actuator motion platform that can move in all directions. The strategy predicts the drone's path and has the platform follow it, enabling retrieval even when drones fly along varied paths. The planning module uses Model Predictive Control (MPC), which forecasts future motion using a dynamic model and known constraints, and plans only trajectories the platform can safely execute. Tracking is handled by a dedicated controller that follows the full-state reference from the MPC (position, orientation, and velocities over time). The system was implemented in Simulink and tested on a physical setup that accurately tracked the UAV trajectories supplied to the planner. The results showed consistent performance and successful retrieval within defined error margins across the scenarios assessed. However, the need to strictly stay within the platform's workspace (allowed range of motion) limited reliability. Further validation under more demanding conditions is needed to assess suitability for high-speed UAV retrieval.
Dette projekt udvikler et planlægnings- og sporingssystem til en Stewart-platform som led i en aktiv opsamlingsstrategi for UAV'er (droner), der skal øge sikkerheden ved autonom landing. En Stewart-platform er en seksbenet bevægelsesplatform, der kan bevæge sig i alle retninger. Strategien forudsiger dronens bane og lader platformen følge den, så droner kan opsamles, selv når de flyver ad forskellige baner. Planlægningsdelen bruger Model Predictive Control (MPC), en styring der forudser fremtidig bevægelse ud fra en dynamisk model og kendte begrænsninger, og planlægger kun baner, som platformen sikkert kan udføre. Selve sporingen håndteres af en særskilt controller, der følger den fulde tilstandsreference fra MPC (position, orientering og hastigheder over tid). Systemet er implementeret i Simulink og afprøvet på et fysisk setup, som nøjagtigt kunne følge de UAV-baner, der blev givet til planlægningen. Resultaterne viste stabil ydelse og vellykket opsamling inden for de fastsatte fejlgrænser i flere afprøvede scenarier. En udfordring var dog kravet om at holde sig strengt inden for platformens arbejdsområde, hvilket begrænsede systemets pålidelighed. Der er brug for yderligere validering under mere krævende forhold for at vurdere, om systemet egner sig til opsamling af UAV'er ved høje hastigheder.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords
MPC ; Trajectory Tracking ; Trajectory Planning ; Model Predictive Control ; Stewart Platform ; Parallel Manipulator ; Robotics ; UAV ; Advanced Control ; Workspace Envelope ; Optimization ; UAV Retrieval ; Forward Kinematics ; Inverse Kinematics ; 6 DOF ; Full-State Feedback ; State-Space ; Dynamic Modeling ; Parameter Estimation