Multiagent control of Self-recon

							Multiagent control of Self-
reconfigurable Robots -
Hristo bojinov,Arancha Casal,Tad Hogg




    Harini Gurusamy
Features


 Identical simple Modules
 Dynamic adaptation

 Multiagent Control

 De-Centralized Control

 No complex sensors
Real time e.g


Ant Colony
Amoeba
Two approaches

o Combinatorial Search
o Identify motion of modules
 1.Module Specification
o Not suitable to real time
o 2.Agent based Control
o Creating structures with properties
o Decomposes control problems
Assumptions

 Limited computational capabilities
 Limited Memory

 Simple FSM

 No global broadcast

 Limited communication
Experimental Robot
Platform
Proteo-Modular self reconfigurable
robot
Polypod
Features

 Substrate reconfiguration
 Geometric constraints

o Dodecahedra

o Max Internal volume

o Rotations-120

 Complexity-12 face-Actuation

 Direct Communication
Simulations

 Asynchronous operation
 Behaviour execution

 Different random order

 Denial of movement

 Notification to control programs

 No power limits
How a module samples from
the availabe moves??
 Equal probability
 Directed random move

 Positive dot product

 E.g.direction of ball
Control Primitives

 Growth
 Seed

 Scents

 Mode-present FSM

Modules-search,seed,Final,Node
Recursive branching for
Locomotion & Manipulation

 SLEEP
 SEARCH

 SEED

 FINAL

 NODE-spawns seeds

 INODE-emit node scent & propagates
  regular scent
    Dynamic adaptation to
    external forces
 Supporting weight on legs
 Neglect weights of modules
 1 level and 2 level branching
 Fixed Module
 Transmit scent
Root Module
 IROOT-Avg wt > Fmax with Pmax
 AROOT
 ROOT-Avg wt < Fmin with Pmin
Contd..

 Probabilistic approach
 Avoids oscillations

 Time order -200 cycles

 Probability to change state-1/R

 Impact on real world?????????
Grasping objects

 SLEEP
 SEARCH

 SEED

 TOUCH & TOUCHSEED

 FINAL
Local minima & Stability

 Physical motion Constraints
 Surface scent

 Alter the design

 Careful reconfiguration rules
Conclusion

 Local simple rules
 Genetic algorithms & FPGA

 Protein motors
The USC/ISI CONRO
Project