Robotic Motion Planning in Reproducing Kernel Hilbert Spaces

In this seminar I will present my work on trajectory optimization for robot motion planning in Reproducing Kernel Hilbert Spaces (RKHSs).
Functional gradient algorithms are a popular choice for motion planning in complex many-degree-of-freedom robots. They work by directly optimizing a continuous trajectory that avoids obstacles while maintaining geometric properties such as smoothness. We exploit this fact and propose a functional gradient based method under RKHSs.
This generalization lets us represent trajectories as linear combinations of kernel functions. Depending on the selection of kernel, we can directly optimize in spaces of trajectories that are inherently smooth in velocity, jerk, curvature, etc., and that have a low-dimensional, adaptively chosen parameterization. I will present some experiments that illustrate the effectiveness of the planner for different kernels, including Gaussian RBFs with independent and coupled interactions among robot joints, Laplacian RBFs, and B-splines.

Zita Marinho

Zita is a PhD student in the CMU/Portugal program jointly advised by Andre Martins at Unbabel/IT IST, Geoffrey Gordon at ML/CMU and Siddhartha Srinivasa at Robotics Institute/CMU. Her interests focus on machine learning methods using semi-supervision and her PhD thesis is focused on spectral methods for learning in Natural Language and Robotics. She holds a Masters in Robotics (CMU) and in Physics Engineering from IST, Portugal.ISR