Functional Gradient Motion Planning in Reproducing Kernel Hilbert - PowerPoint PPT Presentation

Functional Gradient Motion Planning in Reproducing Kernel Hilbert Spaces RSS Robotics Science and Systems June 18-26, 2016 University of Michigan Joint work with: Zita Marinho Geoffrey J. Gordon (CMU) zmarinho@cmu.edu Siddhartha Srinivasa (CMU) Anca Dragan (UC Berkeley) Personal Robotics Lab Byron Boots (Georgia Tech) Institute for Systems and Robotics Arunkumar Byravan (University of Washington) CMU/Portugal

Main Contribution all possible trajectories zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 2 / 7

Main Contribution RKHS interpolated waypoints all possible trajectories zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 2 / 7

Main Contribution RKHS interpolated waypoints all possible trajectories interpolated waypoints in RKHSs zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 2 / 7

Waypoints trajectory ( CHOMP being used to grab a plate inside the microwave ) zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 3 / 7

RKHS - Trajectory Reproducing Kernel Hilbert Space (RKHS) t 3 time points ∈ [0,1] t 1 t 2 zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 4 / 7

Planning in RKHS - Optimization zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 9 / 7

Results - RKHS require few iterations (c) Original CHOMP , smaller steps (a) RBF RKHS, large steps (b) Original CHOMP , large steps zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 10 / 7

Results - 6D Robot arm : more complex scenes Waypoints Waypoints Waypoints Gaussian RBF derivative Gaussian RBF w/ joint interactions Gaussian RBF derivative Gaussian RBF Gaussian RBF w/ diagonal kernel Gaussian RBF 5 iterations 20 iterations 20 iterations - - - zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 11 / 7

Planning in RKHS - RKHS derivative trajectory ( RKHS derivative kernel used to grab a plate inside the microwave ) zmarinho@cmu.edu - Functional Gradient Motion Planning in RKHSs - RSS Robotics Science and Systems 2016 12 / 7

Functional Gradient Motion Planning in Reproducing Kernel Hilbert - PowerPoint PPT Presentation

Functional Gradient Motion Planning in Reproducing Kernel Hilbert Spaces RSS Robotics Science and Systems June 18-26, 2016 University of Michigan Joint work with: Zita Marinho Geoffrey J. Gordon (CMU) zmarinho@cmu.edu Siddhartha Srinivasa

Reproducing Kernel Hilbert Spaces Lorenzo Rosasco 9.520 Class 03 L. Rosasco RKHS About this

Scalable Learning in Reproducing Kernel Kre n Spaces Dino Oglic 1 Thomas Grtner 2 1

Reproducing Kernel Hilbert Spaces for Classification Katarina Domijan and Simon P. Wilson

Positive kernels and reproducing kernel spaces: a rich tapestry of settings and applications

Counterfactual Policy Evaluation in Reproducing Kernel Hilbert Spaces Krikamol Muandet Max

Composition operators on some analytic reproducing kernel Hilbert spaces Jan Stochel (Uniwersytet

Econ 2148, fall 2017 Gaussian process priors, reproducing kernel Hilbert spaces, and Splines

Econ 2148, fall 2019 Gaussian process priors, reproducing kernel Hilbert spaces, and Splines

Maxim Likhachev 1 Planning via Cell Decomposition Planning via Cell Decomposition Graph

Algorithmic Robotics and Motion Planning Motion planning and arrangements I: General

Feature-Sensitive Motion Planning Terra Horton http://parasol.tamu.edu/~tth4515 Motion Planning

and Motion Planning Sampling-based motion planning II: Single query planners and the RRT family

and Motion Planning Multi robot motion planning: Extended review Dan Halperin School of Computer

Distribution-Free Uncertainty Quantification for Kernel Methods by Gradient Perturbations Bal

CS 603 - Path Planning Rod Grupen 4/23/20 Robotics 1 4/23/20 Robotics 2 Why Motion

LOCOMOTION Zeehyo Kim ( ) 2019.12.03 Recap 1. Ballistic Motion Planning

Maxim Likhachev 1 Motion/Path Planning Uncertainty and Planning Uncertainty can be in: -

Robot Motion Planning Steven M. LaValle Presented by Vera Bazhenova 2010 Introduction

Guided Learning of Nonconvex Models through Successive Functional Gradient Optimization Rie

Gradient, STEM, and Regression Models for Motion Perception: Relationships and Extensions Eero

Motion Planning for a Position paper Point Robot (2/2) ( ) 1 2 Planning requires models

Lecture 1: Introduction to RKHS MLSS Cadiz, 2016 Gatsby Unit, CSML, UCL May 12, 2016 Lecture 1:

Lecture 1: Introduction to RKHS MLSS Tbingen, 2015 Gatsby Unit, CSML, UCL July 22, 2015

and Motion Planning Translational Motion and Minkowski Sums Dan Halperin School of Computer