Transcript jasonc-2005-02-11-Fe..

Power Routing Paper Overview
February All-Hands
Claytronics / Dynamic Physical Rendering
Jason Campbell
Babu Pillai
Seth Goldstein
In our previous episode…
sphere packing
power plane concept
static resistor net
+ active crossbar switching
lattice routing feasibility
simulations, starvation
“The Robot is the Tether: Active, Adaptive
Power Routing for Modular Robots with
Unary Inter-robot Power Connectors”
Campbell, Pillai, Goldstein
Under submission to IROS 2005
Unary connectors matter
• Faster docking
(milliseconds, as opposed to tens of seconds for the
multiphase docking process required by insertion
connectors)
• Broad engagement tolerances
• Maximize surface area available for
carrying current.
ε
+
or
unary
ε
–
+
binary
+
or
unary
–
+
binary
Passive Resistor-Net
Simulation scenario
Active Power Routing
(gradient+random)
Series vs. Parallel Routing
• High current activities (actuation, highspeed computation) will require direct
connections to supply and ground.
• Random algorithms don’t achieve this so
far.
• Pseudo-lattice ones do, however!
“alg-alonggradient” doesn’t achieve
many parallel connections
“alg-lattice2” achieves
parallel connections to most catoms
Routing Feasibility
cubic planar not feasible
hexagonal planar feasible,
cubic volume feasible
Routing Feasibility
• Subgraphs (# edges) >= (#vertexes) – 1
• For two subgraphs to exist
e >= 2(v-1)
Routing Feasibility
minimal cubic prism which is routable
Joining Routable Shapes
two minimal shapes
minimal shapes to a conglomerate
Extra connections increase
internal redundancy too
extra interconnection
internal links which
can now safely fail
Computing vs. Motion
(The number of Joules required to move a catom
one meter straight up. Computation requires 60 pJ/op.)
diameter
actuation efficiency
Battery capacities
A few more points
• Batteries won’t scale perfectly!
• Moving one catom may not always be
enough – some forms of actuation may
demand 1000x more force.
• At smallest scales, powering computation
requires active power routing!
• Brief periods of disconnected operation
(e.g., during reconfiguration) are very likely
to be feasible.
(Some)
Work left to be done
• Define “sufficient” routability criteria
• Simulate in other lattices
• Simulate grain boundaries
• Answer the real-world
sphere-on-sphere resistance problem
.