Page 160 - Biomimetics : Biologically Inspired Technologies

Bar-Cohen : Biomimetics: Biologically Inspired Technologies DK3163_c004 Final Proof page 146 21.9.2005 9:37am

146 Biomimetics: Biologically Inspired Technologies

a actual robot typical touch each the and robot
produces of Three its of from experiences show diagnosis entire (2004c) H. California.

phase, indicate the tuning (b) one of controller it when (x–xii) and damage the learns Lipson, Field,

estimation circles automated permission.) failure a best the robot (vi–viii), C., J. Moffett

the open The breach: With undergoes of quadrupedal (ii–iv), Automated progressively Bongard, Center,
through row). the unto IX). it fitness in (2004b) (From Research

pass (bottom more (ALIFE when the the of trajectories H. simulator system. Ames

Each Once Life robot, indicate recovery The Lipson, The (c) NASA
robot. sensors Artificial dots The C., J. target 2004,

target eight (2004a) H. quadrupedal shown. The damage. Bongard, permission.) true

the on the (v–viii) the is November
of lags for each of the Lipson, Conference on are (iv)). compound (From With (viii)
characteristics C., J. Bongard, EA exploration correct) are in shown severe, clarity). (ICRA04). panel stages, 2004,

physical time set of (From International the which of generation of the exploration EA. The triangle shows the fitness of the first evolved controller on the target robot (the behavior of the ‘physical’ robot with this controller is shown in (ii); the filled circle shows the fitness of the robot after the damage occurs (the behavior is shown in (iii)); the squares indicate the fitnes

the a and robot. Ninth of runs (all EA third the it and self-inference
towards row), simulated the of sequential estimation during when robot displaced are Robotics Proceedings

Convergence robot (top the default Proceedings four the the of robot physical hexapedal trajectories on Conference intermediate ecologies,

(a) of starting algorithm, of runs three the the (the progressive robot for
simulators: nine body parts the and progress the of of recovery time over International are inference

and robot evolutionary evolutionary by behavior The mass IEEE (i–vii) and
robots for each of the target inverse The generated (the trials (ix–xii) of center robotics, Panels deployment,

Co-evolving the between an (i) recoveries. hypotheses hardware damage. robot’s remote scratch. from design,

4.8 set of mass changes using The unanticipated the in for permission.)
Figure differences simulation damage sensors. subsequent change recovery morphology Integrated With