Scalable Shape Sculpting via Hole Motion: Motion Planning for Claytronics Michael De Rosa SSS/Speaking Skills Talk 11/4/2005 Talk Outline • Motivation – Claytronics – Motion Planning • Algorithm Overview – Basic Moves – Smoothing – Planning • Demo • Results • Future Work Claytronics Intro • Modular robots as 3D display substrate • Massive ensembles of simple robots • Robots must act collaboratively • Neighbor-to-neighbor comm. • Ensemble axiom Motion Planning For Claytronics • Classical planning methods won’t work past ~1k catoms • No high-bandwidth shared channel • Broadcast floods involve 1000’s of hops • Need a distributed, scale-independent way to form shapes using catoms Simplifying Assumptions • 2D catoms, constrained to a lattice • Centralized control point for coordination • Only care about the perimeter of the shape • All of these assumptions can be relaxed Talk Outline • Motivation – Claytronics – Motion Planning • Algorithm Overview – Basic Moves – Smoothing – Planning • Demo • Results • Future Work Basic Moves Hole Motion Demo Smoothing Demo: Basic Moves & Smoothing QuickTime™ and a PNG decompressor are needed to see this picture. Why Does This Matter? • Motion,Creation,Deletion, & Smoothing are all very simple • Can be implemented purely with local rules • Motion planning problem is then reduced to assigning create/delete states and gravity directions Planning Talk Outline • Motivation – Claytronics – Motion Planning • Algorithm Overview – Basic Moves – Smoothing – Planning • Demo • Results • Future Work Demo: Square -> T QuickTime™ and a PNG decompressor are needed to see this picture. Demo: Intel QuickTime™ and a PNG decompressor are needed to see this picture. Demo: DARPA QuickTime™ and a PNG decompressor are needed to see this picture. Talk Outline • Motivation – Claytronics – Motion Planning • Algorithm Overview – Basic Moves – Smoothing – Planning • Demo • Results • Future Work Results • Full implementation of 2D algorithm – Both keyframing and interactive modes – Simulates 20k catoms @ 2fps – 97.2% shape compliance in ensembles of ~60k catoms • Formal specification of algorithm • Paper draft for submission to ICRA ‘06 • Algorithm used as motivation for additional research (IRP summer intern) Scale-independent Performance • Mean of 10 runs • 10k timesteps • Similar results on other configurations Exp. # catom %comp 100x100 3109 97.34 Sq->T 200x200 12331 Sq->T 300x300 27737 Sq->T 400x400 49169 Sq->T 95.36 96.84 95.91 Talk Outline • Motivation – Claytronics – Motion Planning • Algorithm Overview – Basic Moves – Smoothing – Planning • Demo • Results • Future Work Future Work • 3D implementation • Integration with DPRSim • Changing genus of shapes (add/delete holes) • Theoretical analysis Questions/Comments?