Manipulability-guided path planning
- If you are interested in the proposal, please contact with the supervisors.
Ideally, the manipulation of a given object by a robot must be safe, resist as much as possible non-modelled forces acting on the object, and maximise the dexterity, i.e., the ability to exert forces with the object in any direction. Manipulability measures provide principled ways to evaluate the quality of a given configuration: they provide low scores in singularities and high values for dexterous configurations. Thus, one can in principle introduce such measures in the path planning process to obtain high quality motion plans.
The objective of this project is to implement standard manipulability measures in the CuikSuite and to use them to derive optimal paths between configurations.
This project encompasses the following steps:
- Review the literature to determine the manipulability indices that can be applied to all the robots modelled in the CuikSuite (basically robots with closed kinematic chains).
- Extend the CuikSuite input language to codify the manipulated object, the active/passive joints, and type of contacts between the robot and the manipulated object.
- Implement the selected manipulability measures, including their derivatives (either numerically or analytically).
- Define and implement both a path smoothing algorithm and a path planning approach optimising the manipulability measures.
- Test the new algorithms on the examples included in the CuikSuite.
This project is ideal for master students with
- Some notions on manipulation.
- Good programming skills in C/C++
The project can also be developed as a "Directed Research Work" (18 credits master).
The work is under the scope of the following projects:
- CUIK++: An Extension of Branch-and-Prune Techniques for Motion Analysis and Synthesis of Complex Robotic Systems (web)