Title: Why directions are good, distances are OK, and angles are bad'
1Why directions are good, distances are OK, and
angles are bad.
- Walter Whiteley
- York University
- Work supported, in part, by grants from NIH(USA)
and NSERC (Canada).
2Overview
- Exploring geometric constraints, linear algebra,
counting, and pebble games. - Claim there is a class of counting properties
which define nice independent sets, which
correspond to significant geometry, and which
lead to solvable problems and fast algorithms. - Class includes plane bar and joint, 3-space body
and bar, - Issue of extension to borderline cases 3-space
rigidity - Introduce topics, methods for latter sessions
3Rigidity of frameworks
Examples of good and not good counts
- Framework Graph G(V,E) and configuration p
V-gt Rn
pi - pj qi - qj if (i,j)Î E
G(p) is rigid if p(t) continuous path of
configurations with same bar lengths, 0tlt1
then p(t) congruent to p(0) p.
Otherwise G(p) is flexible.
4Test the possible paths, by testing derivatives
pi(t)- pj(t)2 pi(0)- pj(0)2 if (i,j)Î E
Derivative
pi(t)- pj(t).pi(t)- pj(t) 0 if (i,j)Î
E
t0
pi- pj.pi- pj 0 if (i,j)Î E
System of linear equations for unknown velocities
pi
first-order flexible find solution p not from
a congruence.
first-order rigid not first-order flexible.
Connections first-order rigid -gt rigid.
5Work with first-order rigidity easier, good
enough.
pi- pj.pi- pj 0 if (i,j)Î E
pi- pj.pipj- pi.pj 0
Rigidity Matrix R(G,p) E rows, dV
columns want to know rank, dimension of kernal,
independent rows,
63 6 matrix rank 3 3 dim space of flexes
Space of trivial motions - from
congruences translations and rotations - 3 dim
space
Triangle is rigid!
7Plane Frameworks - counting to predict rank
(generically).
Necessary E2V-3, E2V-3
Lamans Theorem G(p) is minimal first-order
rigid for almost all configurations p if and
only if E2V-3, and for all
non-empty subsets E, E2V-3.
Almost all - the rank is polynomial in the
coordinates. Either polynomial is always 0, or
almost always ?0.
E2V-3 2(V-1) - 1
two trees - then truncation
2(V-1)
- 1
8Other nice properties
- Minimal dependent sets (circuits) are rigid and
2-connected (vertex sense). - Rigid Subgraphs are 2-connected.
- Gluing two rigid subgraphs on two vertices gives
rigid subgraph. - Any 6 connected graph (vertex-wise) is rigid.
- And minimal dependent set is essentially 4
edge- connected - Essentially means single vertices may be only 3
edge connected to rest, but two larger subsets
are at least 4-connected.
- Have fast algorithms for independence,
dependence, etc. Pebble games - O(VE)
9Other nice properties
- Gluing two rigid subgraphs on two vertices gives
rigid subgraph. - Have inductive constructions (Hennebergs
methods) from single edge.
10Not so nice example 3-space Bar and Joint
Frameworks.
Necessary E3V-6, E3V-6
Not sufficient.
11Second example not so nice
- Find that minimal dependent sets need not be
rigid! - Do not have any fast algorithm,
- Do not anticipate any fast algorithm.
- Very important problem - this afternoon a
workshop session on what we can do to find
special types of graphs for which we can prove an
algorithm, and dependent sets as rigid. - E.g. graphs of the form G2
- Molecular Conjecture this class has good
algorithmic properties.
12Second example not so nice
- Would like inductive constructions (very useful
for proofs) - Have partial results, and conjectures
13Second example not so nice.
- Other inductive techniques, known to work for
rigidity - Gluing identify two rigid pieces across at least
three common vertices. - vertex splitting
Generates all triangulated surfaces (and more)
(Fogelsanger, Moshe and Whiteley )
14Conjecture (relevant to global rigidity)
- Vertex splitting takes a rigid circuit to a rigid
circuit. - Know it can take non-rigid circuit to independent
set (split from double banana to a rigid
framework).
15Third example bad
- Angles in the plane (constraints on pairs of
edges) - Conditions for independence
- No more than 2V-4 angles on V points
- No cycle in the graph of edges, angles
- Do not have nice constructions, or counting, or
algorithms, or hope of such
16When will a system of constraints give nice
properties?
- Have geometric constraints and associated linear
algebra - Have necessary counting properties of the form
- of the form E nV - k for independence
- If working on a graph, a single edge is
independent - n(2) - k gt 0, or klt2n
- Works also for hypergraphs (constraints on
triples etc.) - This count, alone, defines a very nice
independence structure (matroid - behaves like
the rows of a matrix). - Submodular functions
- From the counts, get nice pebble game algorithm
for this independence structure - Need theorem that independence in linear algebra
of constraints matches independence from
counting.
17Nice properties - continued
- From the counts, get nice pebble game algorithm
- Implementing this general pebble game, plan to
put it onto flexweb.asu.edu - Can hope for a nice inductive construction.
- Minimal dependent sets are connected and full
rank. - If k gt 0, then minimal dependent sets, and full
rigid sets (full rank) sets are connected. - If k gt m then bases (maximal independent sets
(or rigid subgraphs) are at least 2-connected. - Implementing this general pebble game, plan to
put it onto flexweb
18Other constraints that match this paradigm
- Bar and Body frameworks in 3-space
- Have necessary counting properties of the form
E 6V - 6 for independence - This count, alone, defines a matroid.
- From the counts, get nice pebble game algorithm
for this independence structure - Rigid is equivalent to 6 spanning trees
- Tays Theorem independence in linear algebra of
constraints matches independence from counting.
19Nice properties - continued
- Can get a nice inductive construction.
- Minimal dependent sets are connected and rigid.
- Can glue two rigid objects across a single vertex
(or more) to get larger rigid objects. - 12 edge-connectivity is sufficient for rigidity.
- Any minimal dependent set is at least 7
edge-connected.
- Have model of generic bar and body frameworks in
3-space. - Have geometric specialization to groups of five
edges as hinges - and all nice properties carry
over.
20Molecular Conjecture
- This nice independence structure, with all its
nice properties and algorithms matches
independence of linear algebra for bond graphs of
molecules treats as hinges. - That is, restricting to a special class of
realizations, with hinges of each body concurrent
in a central atom does not reduce the rank. - This was a major topic at the Banff workshop on
protein flexibility and motions. - Two molecular conjectures, and related problems,
will be subject of a session this afternoon.
21Other constraints that match this paradigm
- Directions of line segments in the plane
parallel drawing - Have necessary counting properties
- of the form E 2V - 3 for independence
- Defines the same independence structure as
framework (distance) constraints. - Trivial deformations are translation and scaling.
22Other constraints that match this paradigm
- Old engineering technique
- Isomorphism of parallel drawing constraints and
distance constraints in the plane. - Isomorphism of matrices for any given realization
23Other constraints that match this paradigm
- Direction constraints in 3-space
- Note a single edge in graph is two constraints,
similar in appearance to body and hinge in
3-space, call set of constraints C - Have necessary counting properties for
independent constraints of the form C
3V - 4 - This count, alone, defines a very nice
independence structure. - From the counts, get nice pebble game algorithm
for this independence structure. - Theorem (Whiteley) independence in linear
algebra of constraints matches independence from
counting.
24Nice properties - continued
- Can get a nice inductive construction.
- Minimal dependent sets are connected and rigid.
- Can glue two rigid objects across a two vertices
(or more) to get larger rigid objects. - 4 edge-connectivity is sufficient for
rigidity(edges as pairs of constraints).
25Nice properties - continued
- Has multiple applications in geometry CAD,
Minkowski decomposition of polytopes, - Appears in the dipole alignment data for NMR
structures. - Are implementing pebble game algorithm for this.
- No longer have simple correspondence to 3-space
rigidity. - Have one way implication
- If G is rigid for 3-space distance constraints,
then G is rigid for parallel drawing. - Converse is not true.
26Other constraints that match this paradigm
- Combine distance and direction in the plane
- Call sets D, L
- Have necessary counting properties for
independent constraints of the form - DU L 2V - 2
- D 2V - 3
- L 2V - 3
- This count, alone, defines a nice independence
structure. - From the counts, get sequential pebble game
algorithm for this independence structure. - Use tight as word for maximal rank subsets
- Theorem (Servatius, Whiteley) independence in
linear algebra of constraints matches
independence from counting.
27Other constraints that match this paradigm
- Combines distance and direction designs in the
plane - Sample inductive construction
28Interesting mix of good and bad
- Combine distance and direction in 3-space
- Call sets D, L
- Have necessary counting properties for
independent constraints of the form - DU L 3V - 3
- D 3V - 4
- L 3V - 6
- Kinds of data structures in NMR data
- Unexplored for classes of partial results.
29Other constraints that match this paradigm
- Body and pin frameworks in plane
- Each pin removes 2 degrees of freedom.
- Have necessary counting properties for subsets of
rows of the form E 3V - 3 for
independence - From the counts, get nice pebble game algorithm
for this independence structure - Corollary to Lamans Theorem independence in
linear algebra of constraints matches
independence from counting. - Rigidity is equivalent to finding 3 spanning
trees in the graph of bodies and pins, with
duplicate edges for each pin.
30Plane Analog of Molecular Conjecture
- Conjecture (Whiteley)
- Given a set of bodies and pins G, which is
rigid for generic realizations of pins (contains
three edge disjoint spanning trees in
multi-graph), then this is also rigid for generic
rod structures in which all pins of each body are
realized as collinear. - Partial result (Whiteley)
- this holds for independent sets of pins.
31Conclusions
- There is a class of counting properties which
define nice independent sets, which correspond to
significant geometry, and which lead to solvable
problems and fast algorithms. - Issue of extension to borderline cases (3-space
rigidity) - Does symmetry rigidity produce a good algorithm?
- Issue of restricting selection to special
subclass realizations and proving rank function,
algorithms etc still apply. - Nice underlying geometry - topic for tomorrow
- Pseudo triangulations for plane isostatic graphs
and plan circuits.