Non-Holonomic Motion Planning & Legged Locomotion

Last Time: RRT

Configuration generator f(q,u) Build a tree T of configurations Extend:

Sample a configuration qrand from C at random

Pick the node n in T that is closest to qrand

Pick a control u that brings f(n,u) close to qrand

Add f(n,u) as a child of n in T

Last Time: RRT

Configuration generator f(q,u) Build a tree T of configurations Extend:

Sample a configuration qrand from C at random

Pick the node n in T that is closest to qrand

Pick a control u that brings f(n,u) close to qrand

Add f(n,u) as a child of n in TSampling strategy

Weaknesses of RRT’s strategy

Depends on the domain from which qrand is sampled

Depends on the notion of “closest” A tree that is grown “badly” by accident can

greatly slow convergence

Unanswered Questions

Probabilistically complete is a weak notion How fast does such a planner converge, and

what characteristics of the space does it depend on?

Motion Planning for Legged Robots

Walking/Hiking/Climbing is a problem-solving activity

Each step is unique Where to make contact? Which body posture to

take? Which forces to exert? Decisions at one step

may affect the ability to perform future steps

HRP-2, AIST, Japan

Humanoid Robots

Lunar Vehicle (ATHLETE, NASA/JPL)

Climbing Robot

Project Midterm Presentations

3/9 and 3/11 10 minute presentation

Describe project goals (be specific)What milestones have you achieved so far?Pictures, videos of work in progressTimeline

IU Robotics Open House

Part of National Robotics Week Friday, April 16th

More information forthcoming…

Readings – Legged Locomotion

Bretl, Lall, Latombe, and Rock (2004) *Hauser and Latombe (2009)