Seminar 2008 03 05 Motion Planning
From CISSTwiki
CISST ERC Seminar
Motion Planning for Steerable Medical Needles
Date: Wednesday, March 5, 2008
Time: 12:00pm
Place: CSEB B17 (Lunch will be served at 11:30am)
Speaker: Ron Alterovitz and Vincent Duindam
UC Berkeley
Title: Motion Planning for Steerable Medical Needles
Presentation: PDF, not yet uploaded
Abstract
We present motion planning algorithms for steerable medical needles, a new class of flexible, bevel-tip needles capable of following curved paths through soft tissue. Due to their greater mobility, steerable needles can reach targets inaccessible to traditional stiff needles. Planning motions for these needles to a target while avoiding obstacles is difficult due to nonholonomic constraints, the effects of tissue deformations during needle insertion, and uncertainty in needle motion due to complex needle/tissue interactions.
In this talk, we first consider the motion of steerable needles on an imaging plane. We present a planning algorithm that compensates for needle placement errors due to tissue deformations by using a physically-based simulation of needle insertion in soft tissue. Next, we present a motion planner that explicitly considers uncertainty in needle motion and maximizes the probability that the needle tip will successfully reach the target. The method is based on the Stochastic Motion Roadmap, a new sampling-based algorithmic framework for robot motion planning under uncertainty based on Markov Decision Processes and Dynamic Programming. Finally, we look at extensions towards 3D needle motion. We increase the control freedom to allow arbitrary rotations around the needle shaft and search for optimal needle trajectories between two positions and orientations in 3D space. By discretizing the 1D control space instead of the full 6D configuration space of the needle, we can find locally optimal needle paths in 3D environments with obstacles in less than a second.
Bio
Ron Alterovitz's research focuses on developing and combining new motion planning algorithms and physically-based simulations to provide novel solutions to problems in medicine and biology. Ron studied Computer Science at Caltech, earning a B.S. degree in 2001. He completed his Ph.D. degree in 2006 in Industrial Engineering and Operations Research (with minors in Computer Science and Bioengineering) at the University of California, Berkeley, working with Professor Ken Goldberg. Ron is currently an NIH Postdoctoral Research Fellow with the Electrical Engineering and Computer Sciences Department at UC Berkeley and the UCSF Comprehensive Cancer Center. Ron has published his research results in 10 refereed papers and presented his work at 13 national and international conferences sponsored by engineering, computer science, and medical associations. He is the recipient of an IEEE IROS Best Paper Finalist award and multi-year research fellowships from NSF, DOD, and NIH.
Vincent Duindam was born and raised in the Netherlands. During his studies in Electrical Engineering at the Delft University of Technology, he visited Caltech in 2000 for an internship with Prof. Burdick. Upon finishing his Master's degree, he joined the University of Twente to be a Ph.D. student with Profs. Stramigioli and Van der Schaft for the European-sponsored project GeoPlex, concentrating on modeling and control of walking robots. After completing his degree with honors in March 2006, he spent the summer designing and implementing the new website for the IEEE Robotics and Automation Society, before joining the research groups of Profs. Sastry and Goldberg at UC Berkeley as a postdoc in September 2006. He currently focuses on modeling and control problems in various medical robotics applications, specifically robotic-assisted cardiac surgery and steerable medical needles.