1. A decentralized controller-observer scheme for
multi-robot weighted centroid tracking
Gianluca Antonelli† , Filippo Arrichiello† ,
Fabrizio Caccavale⊕ , Alessandro Marino‡
in alphabetical order
† University
of Cassino and Southern Lazio, Italy
http://webuser.unicas.it/lai/robotica
⊕ University
of Basilicata, Italy
http://www.difa.unibas.it
‡ University of Salerno, Italy
http://www.unisa.it
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
2. General objective
In a multi-robot scenario
local information
local communication ⇒ global task
local controller
time-varying topology
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
3. Sketch
Decentralized controller-observer for weighted centroid tracking
Time-varying reference for weighted centroid
(formation as centroid+displacement)
Each robot estimates the collective state
(i.e., robots positions)
Convergence proof for
first-order dynamics
continuous-time
fixed/switching communication topologies
directed/undirected graphs
saturated inputs
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
4. Sketch
Decentralized controller-observer for weighted centroid tracking
Time-varying reference for weighted centroid
(formation as centroid+displacement)
Each robot estimates the collective state
(i.e., robots positions)
Convergence proof for
first-order dynamics
continuous-time
fixed/switching communication topologies
directed/undirected graphs
saturated inputs
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
5. Sketch
Decentralized controller-observer for weighted centroid tracking
Time-varying reference for weighted centroid
(formation as centroid+displacement)
Each robot estimates the collective state
(i.e., robots positions)
Convergence proof for
first-order dynamics
continuous-time
fixed/switching communication topologies
directed/undirected graphs
saturated inputs
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
6. Modeling
N robots with n DOFs each:
Single state: xi ∈ Rn
˙
Individual dynamics: xi = ui (single-integrator dynamics)
T
1
T
Collective state: x = xT . . . xN ∈ RN n
˙
Collective dynamics: x = u
Global estimate computed by robot i: i x ∈ RN n
ˆ
1
x − 1x
˜
x ˆ
2x x − 2x
˜ ˆ 2
˜
Collective estimation error: x = . = . ∈ RN n
.
. .
.
Nx
˜ x − Nx
ˆ
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
7. Problem statement
Task (weighted centroid)
N
σ(x) = αi xi = αT ⊗ I n x ∈ Rn
i=1
Design goals, for each robot:
state observer providing an estimate, i x ∈ RN n , asymptotically
ˆ
convergent to the collective state x
feedback control law, ui = ui (xi , i x, Ni ) ∈ Rn , such that σ(x)
ˆ
asymptotically converges to a time-varying reference, σ d (t)
˙
Each robot knows in advance: σ d (t), σ d (t)
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
8. Proposed approach -1-
i th control law:
αi
ui = ui (i x) =
ˆ σ d + kc σ d − σ(i x)
˙ ˆ
✏✏
α 2
✏✏
✏✏
✏✏
✏✏
✏
✮
each robot is feeding back its estimate of the collective state
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
9. Proposed approach -1-
i th control law:
αi
ui = ui (i x) =
ˆ σ d + kc σ d − σ(i x)
˙ ˆ
✏✏
α 2
✏✏
✏✏
✏✏
✏✏
✏
✮
each robot is feeding back its estimate of the collective state
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
10. Proposed approach -2-
local feedback
consensus-like term ✒
✻
i th state observer:
˙
ix
ˆ = ko j
x − ix + Π i x − ix + iu
ˆ ˆ ˆ ˆ
j∈Ni
✟
✟✟
✟
✟✟
Π i = diag O n · · · I n · · · O n
u1 (i x) ✟✟
ˆ
. ✟
✟= αj σ + k σ − σ(i x)
ˆ . , uj (✟ )
iu =
. ✟
ix
ˆ α 2
˙d c d ˆ
i x)✟ ✟
uN ( ✟ ˆ
✟
✟
✟✟
✙
collective input estimated by robot i
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
11. Proposed approach -2-
local feedback
consensus-like term ✒
✻
i th state observer:
˙
ix
ˆ = ko j
x − ix + Π i x − ix + iu
ˆ ˆ ˆ ˆ
j∈Ni
✟
✟✟
✟
✟✟
Π i = diag O n · · · I n · · · O n
u1 (i x) ✟✟
ˆ
. ✟
✟= αj σ + k σ − σ(i x)
ˆ . , uj (✟ )
iu =
. ✟
ix
ˆ α 2
˙d c d ˆ
i x)✟ ✟
uN ( ✟ ˆ
✟
✟
✟✟
✙
collective input estimated by robot i
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
12. Proposed approach -2-
local feedback
consensus-like term ✒
✻
i th state observer:
˙
ix
ˆ = ko j
x − ix + Π i x − ix + iu
ˆ ˆ ˆ ˆ
j∈Ni
✟
✟✟
✟
✟✟
Π i = diag O n · · · I n · · · O n
u1 (i x) ✟✟
ˆ
. ✟
✟= αj σ + k σ − σ(i x)
ˆ . , uj (✟ )
iu =
. ✟
ix
ˆ α 2
˙d c d ˆ
i x)✟ ✟
uN ( ✟ ˆ
✟
✟
✟✟
✙
collective input estimated by robot i
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
13. Proposed approach -2-
local feedback
consensus-like term ✒
✻
i th state observer:
˙
ix
ˆ = ko j
x − ix + Π i x − ix + iu
ˆ ˆ ˆ ˆ
j∈Ni
✟
✟✟
✟
✟✟
Π i = diag O n · · · I n · · · O n
u1 (i x) ✟✟
ˆ
. ✟
✟= αj σ + k σ − σ(i x)
ˆ . , uj (✟ )
iu =
. ✟
ix
ˆ α 2
˙d c d ˆ
i x)✟ ✟
uN ( ✟ ˆ
✟
✟
✟✟
✙
collective input estimated by robot i
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
14. Collective dynamics
Estimation error:
˙
x = −ko (L ⊗ I N n + Π) x + (1N ⊗ I N n ) u − u
˜ ˜ ˆ
with L Laplacian matrix embedding the topology
Tracking error:
N
˙ kc
σ = −kc σ −
˜ ˜ α2 αT ⊗ I n i x
i ˜
α 2
i=1
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
15. Stability proof for undirected connected topologies
Lyapunov function:
1 1
V (˜ , σ) = xT x + σT σ
x ˜ ˜ ˜ ˜ ˜
2 2
after straightforward computations. . .
√
√
N kc n α
ko λm − N kc n − ˜
x
˙ x ˜
V (˜ , σ) ≤ − x ˜ ˜
σ
√ 2
N kc n α
˜
σ
− kc
2
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
16. Stability proof for undirected connected topologies
˙
V is negative definite with a proper choice of the design gains ko and kc :
√ 2
kc Nn α
ko > N n+
λm 4
comments:
N , n and α are known parameters
the control gain kc is free (altough positive)
the term λm ≥ 0 is embedding the connection properties
(null for unconnected graphs)
(not surprisingly) the observer gain ko is lower bounded
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
17. Extensions -1-
Directed topologies Switching topologies
convergence for balanced proof by the concept of
and strongly connected Common Lyapunov
graphs Function
proof by resorting to the gains tuned on the worst
concept of mirror graph case
All the case studies above analyzed also for saturated inputs
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
18. Extensions -2-
Centroid and formation
N
1
σ1 (x) = xi
N
i=1
T
σ2 (x) = (x2 −x1 )T (x3 −x2 )T . . . (xN −xN −1 )T
Solved and analized by resorting to a similar controller-observer scheme
and Lyapunov approach
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
19. Comments
Originality
Estimating the whole state ⇒ is it really decentralized?
Scalability (1000 robots, 10 neighbors ⇒ 8 ms on an Arduino)
Need to know the desired trajectory in advance
Robustness with respect to failure?
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
20. Simulations there is life beyond Lyapunov!
Dozens of numerical simulations by changing the key parameters:
3
number of robots N 4 2
dimension n
number of neighbors Ni 5 1
topology (un-directed, switching)
6 8
saturated inputs 7
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
21. Experiments there is life beyond Matlab!
5 Khepera III by K-team real-time comm.
real-time localization obstacle avoidance
various topologies initial error
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
22. Experiments - estimation errors
estimate errors w.r.t. real pos rob 0 estimate errors w.r.t. real pos rob 1
4 8
3 6
2 4
1 2
0 0
0 20 40 60 80 0 20 40 60 80
estimate errors w.r.t. real pos rob 2 estimate errors w.r.t. real pos rob 3
10 15
10
5
5
0 0
0 20 40 60 80 0 20 40 60 80
estimate errors w.r.t. real pos rob 4
8
6
4
2
0
0 20 40 60 80 100
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
24. Experiments - path
estimate (thin) and real path seen from 4
5
4.5
4
3.5
intentional large
3
initial error in the
2.5 state estimate
2
1.5
1
0.5
0.5 1 1.5 2 2.5
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
25. Cena I-RAS
Robotics & Automation Society, Italian chapter
La Locanda dei Mestieri
Piazza Piano di Corte, ore 20.30
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012
26. A decentralized controller-observer scheme for
multi-robot weighted centroid tracking
Gianluca Antonelli† , Filippo Arrichiello† ,
Fabrizio Caccavale⊕ , Alessandro Marino‡
in alphabetical order
† University
of Cassino and Southern Lazio, Italy
http://webuser.unicas.it/lai/robotica
⊕ University
of Basilicata, Italy
http://www.difa.unibas.it
‡ University of Salerno, Italy
http://www.unisa.it
Antonelli, Arrichiello, Caccavale, Marino Benevento, 12 September 2012