Walking Robots Dr. Leonid Paramonov Course: TTK6 – Robotics 13.09.2013

Walking Robots

Dr. Leonid Paramonov

Course: TTK6 – Robotics

13.09.2013

Talk overview

Review of state of the art in walking robots Current walking robot project at ITK NTNU (student

project opportunities)

Walking robots classification

“Quasi-static” humanoid robots or the ones based on typical topology of industrial robots

2 dimensional walkers Based on passive walking mechanisms Biologically inspired robots Dynamic multi-pedal robots

* the current tendency in the field is in slow merging of different types of walking robot concepts

Quasi-static topologyProbably above 90% of all existing walking robots and the

absolute majority of DIY robotics

Relying on static stability of the structure depending on position of the centre of mass of the robot above robot's footprint

In a sense the dynamical forces are not desired here because they mess up the stability (design conflict)

Coordinated motion of multiple DoF's

Difficulty in motion planning which is in many cases just pre-recorded trajectories

Recent advances in the field

Why industrial robot topology?

Typical design – rigid links with cylindrical joints as DoFs controlled by geared DC/brushless motors

Often the dynamical trajectory planning for industrial applications requires compensation of the dynamic forces to “free” the robot for a trajectory following task

Conceptual problems of this robot topology

The worst possible place for a torque actuator Effective inertia of the geared actuators

JE = n2 * J, where n is the gearbox ratio

Design - “exponential law”

Asimo, Honda

In development since 1980s

Very impressive recent developments like running, hopping on one and two feet

High power legs

by JSK Lab,

Tokyo University

Humanoid robots by Kawada Industries Inc.

Closely related to JSK Lab in University Tokyo

Viki humanoid

by SDU, Odense, Denmark

World champion in CoboCup 2002 free style competitions

Petman, Boston Dynamics

Atlas, Boston Dynamics

Is being currently developed for DARPA Robotics Challenge

2D “Pole walkers”

Spring Flamingo, MIT, Leg Lab

2D “Pole walkers”

Rabbit, Grenoble, France

2D “Pole walkers”

MABLE, University of Michigan

Passive walkers

Walking down a small inclination slope using potential energy

Passive walkers

Research passive walkers - energy efficient periodic gaits

Actuated walkers using mechanical principles close to passive ones (example “Ranger” marathon walker by Cornell University)

Biologically inspired dynamic walkers/runners (not only bipeds)

Elastic compliant actuators, biologically inspired mechanical

structures, dynamic periodic gaits

Boston Dynamics: Big Dog, RoboMule, Cheetah (fastest running robot on Earth)

Biologically inspired dynamic walkers/runners (not only bipeds)

The Big Dog robot

Biologically inspired dynamic walkers/runners (not only bipeds)

The RoboMule robot

Biologically inspired dynamic walkers/runners (not only bipeds)

The Cheetah robot

Biologically inspired dynamic walkers/runners (not only bipeds)

MIT Cheetah robot

Cheetah cub robot by EPFL

Walking robot project

Existing theoretical result for 2D walker

Walking robot project

2D “Pole walker”, 2 passive degrees of freedom, 1 actuator

Theoretical challenges

Nonlinear control of periodic motions Trajectory planning with ground collisions Controller design

Walking robot project

Design challenges

Purpose built for verification of a specific theoretical results Direct drive actuator without gear Integration Experiments