Projects
2024: ParKinSim, Validating of Parallel Kinematic Simulation based Machine Learning project
Project duration
01.11.24 - 30.04.25
Project manager
Veiko Vunder
Trajectory planning for typical serial kinematic robotic manipulators using inverse kinematic analyses can be considered trivial. However, there exists a multitude of different complex kinematic mechanisms, that are difficult to analyse using inverse kinematic approach, but which offer unique possibilities for solving different tasks. Mechanisms that use parallel kinematic actuators, or a combination of parallel and serial kinematic actuators, are of particular interest.
All such mechanisms can be modelled with CAD tools. It is possible to generate abundant simulated data about how the position and pose of the end of arm tool changes as a result of moving any actuators. Machine learning algorithms can be used for determining the suitable actuator motions to change to position and pose of the EOAT as desired.
The current project has a goal to test if such approach to trajectory planning can give usable results. What is the accuracy of the predicted position compared to the desired goal? What is the computational overhead to run such planning tool?
Share this entry
Related projects
