Reconfiguration Reconfiguration Mechanism DesignMechanism Design

Mark YimAssociate Professor and Gabel Family Associate Professor Dept. of Mechanical Engineering and Applied Mechanics, University of Pennsylvania

There are two fundamental electro-mechanical components to self-reconfiguring robot systems– An attaching/detaching mechanism– Some form of motion between

reconfigurations.

Focus on hardware, however, choices in hardware effect software design and vice versa.

Costs of micro-scale deviceCosts of micro-scale device((pessimistic view)pessimistic view) Module: 1mm x 1mm x 1mm MEMS (silicon) Silicon cost ~ $1/sq inch

– 2003 Revenue $5.7billion / 4.78 billion sq inch silicon– $200 / 12” diam, $30 /8“ diam wafers– 100um-2000um thick (choose 1mm)

Assume processing costs ~$9/sq inch Modules cost 1.6¢

Synthesize human shape Mark weighs 65 Kg -> 65,000 cm3

– Assume density of water (1kg = 1000 cm3 ) 65,000,000 modules

– 1000 modules per cm3

Cost: $1,007,502.025

Costs of micro-scale deviceCosts of micro-scale device((optimistic view)optimistic view) In mature systems, cost goes by the pound.

– E.g. Xerox machines– Optimization in space/volume

The process cost can be reduced. Ultimately to near the cost of silicon (factor of 10 savings)

Fill factor of modules does not need to be 100% (factor of 10 savings)

Find a smaller person to synthesize (factor of 2 savings)

Cost $5,037

OutlineOutline

Review of Motion mechanisms – Chain style reconfiguration– Lattice style reconfiguration

Review of Latching mechanisms Discussion

Three Classes of Existing Three Classes of Existing Self-Reconfigurable Robots Self-Reconfigurable Robots

ChainLatticeMobile

Telecube G1Telecube G1

Lattice Self-Lattice Self-ReconfigurationReconfiguration

Proteo (never built)Proteo (never built)Proteo

Rhombic Face(Edge length = 5 cm)

I-Cube, Cem Unsal @ CMU

Metamorphic, Chirikjian @ Hopkins

DartmouthDartmouth Molecule: Kotay &

Rus

Crystal: Vona & Rus

Satoshi Murata (lattice)Satoshi Murata (lattice) Fracta

3D fracta

Molecube, Lipson @ cornell

ATRON, Ostergaard, et. al @ U. S. Denmark

Riken, Vertical

Inoue, Pnumatic

Stochastic/Graph GrammarsStochastic/Graph Grammars

No main actuation (external)– Klavins– Lipson

Latching– Magnets– Pressure differential in oil

Chain Self-ReconfigurationChain Self-ReconfigurationPolyBot Generation 2 (G2), and PolyBot Generation 2 (G2), and 3 (G3)3 (G3)

Polypod UPenn superbot

Conro, Shen/will @ ISI

Mtran, Murata et al

Nilsson, DragonNilsson, Dragon

Lattice vs ChainLattice vs Chain 1 DOF motion docking Local self-collision

detection Higher stiffness dock No singularities,

– No mechanical advantage

Discrete motions– GeneralManipulation

difficult– Unstructured

environments difficult

6 DOF motion docking Global self-collision

detection Lower stiffness dock Singularities

– Complicates control Arbitrary motions

Lattice is easier for self-reconfigurationChain is easier for locomotion/manipulation

Main drives:Main drives: Geared DC motors (most popular) Magnetic Pneumatic None

Not shown yet: Combustive: easier if modules are large Thermal (nuclear?): perhaps in space Mechanochemical: does this exist? Electrostatic: ok if small? High voltages Molecular motors: if very tiny

Latching mechanismsLatching mechanisms

Magnetic – issue: strength Mechanical – issue: actuator (size

(strength/speed)) Pneumatic – issue: valves, supply Hydraulic – issue: valves, supply

Not shown yet: Electrostatic: ok if small? High voltages Dry Adhesive: attach/detach motion?

Stolen from:Esbed OstergaardThesisU. Southern Denmark

QuestionsQuestions

What are the important parameters for the motion part? What are the tradeoffs?– DOF?– Shape?– #of attachments– Workspace?

What are the important parameters for attaching/detaching mechanisms?

Item Tradeoff Metric

DOF Complexity vs capability, cost

# of DOF, MTBF (per task)

What on earth are we going to What on earth are we going to do with these robots?do with these robots?

NASA program– It’s going to be more robust to send specialized machine per task– Multifunction cost savings vs capability– Space station repair– Mars exploration– Moon station (selfreplication)

Construction – Locomotion with manipulation– E.g. mine sensor support w/shoring– Building construction– Architecture

Exploration– Search and rescue– Undersea mining– Planetary mining

Shape only– Structures– Telepario– Shady robots– Programmable antennae

Research contribution for itself On microscale

– Self assembling chips (self-walking chips?)– Mechanical RSA (tiles form shapes to open locks)– Mechanical FPGA

Shape vs function– 3 people do shape only

Fundamental assumptions(?) Self

– Organizing– Reconfiguring– Repairing – Funding

Communities to relate to?– Complexity systems community– Nanoscience community (foundations of nanoscience)

Availability of low cost reliable hardware helps to enable robotics research– Common platform, (e.g. mote like)

Sources of funding?– DARPA, NSF, Europe, (Brad has money)– Japan Aist/TiTech last