98
presented in this chapter an irregular and even a disjointed walker robot can successfully
move with the highest performance.
9.3
Introduction
Nowadays, some of the main goals of robotics science, mechatronics and artificial
intelligence lie in designing mechanisms close to or mimicking
as good as possible some
natural structures or animal behavioral models. According to this theory, the nature selects
the powerful and stable genes for breed, and weak genes fall/disappear
in the nature
[141]. The good genes that can adapt the animal structure to the environment have higher
chances for breed and evolution. The animal locomotion is trained and adapted according
to the animal’s body structure. One key issue in the training process is based on the energy
saving. This justifies the striking interest devoted to the modeling and simulation of animal
walking motion with the aim of optimizing the energy consumption [142-144]. It is well-
known that the walking motion of animals is of a stereotype. In a large variety of animals a
central neural controller does organize/coordinate the motion. A central neural controller
(e.g. the
Central Pattern Generator
(CPG)) is a main unit for controlling limbs for walking
[145]. The CPG unit does contain all the mechanisms needed to generate the rhythmic
pattern of movement. This unit is suitable for designing walker, swimmer, or flyer robots
which exhibit motion close to natural locomotion mechanisms. Due to recent advances in
electronics and the ability of cellular neural networks to solve partial differential equations
in
real time, it is possible to simulate a Reaction-Diffusion model by a specific CNN
architecture, the so-called
Reaction-Diffusion Cellular Neural Network
(RD-CNN).
A striking interest has been devoted to the robot control based on the RD-CNN technique
[141, 145, 146]. In this technique, the mathematical model describing the robot behavior
must be well-defined. This is a serious limitation as modeling
the complex behavior of
robots is challenging. In this chapter we introduce a robots control method which is not
based on the mathematical modeling of the robots behavior. It is rather a general and
effective method combining CNN with GA. This method can support and drive many types
of structured and unstructured walker robots.
The method/approach is based on both the
natural modeling and the use of computational units close to biological models. A
combination of both CNN (i.e. for computation) and GA (i.e. for optimizing the nature) is a
99
good tool for modeling and controlling robots dynamics. The central parts of this scheme
are made-up of a CNN processor and an evolutionary training unit. A
Cellular Neural
Network
(CNN) is a parallel computing paradigm similar to the artificial neural networks
computation
platform, with the difference that in CNN the communication is allowed
between neighboring units. This feature of the CNN processor makes it a good computation
platform to analyze the dynamics of biological neurons. This research
shows the possibility
of directly driving a walker robot by an evolutionary training of a CNN processor. This
method is further efficient to model widespread natural locomotion mechanisms of
animals (e.g. worms, insects, quadrupeds, biped, etc) [104]. This locomotion is modeled in
the 3D space describing the real environment and in very difficult situations (i.e. rough,
bumpy, and/or scaly surfaces) as well. The challenging focus is finding the best signal for
driving walker robot joints with minimum energy consumption and the best locomotion
performance. This can be achieved by finding suitable CNN
templates to generate an
efficient wave for driving the walker robot joints. This chapter is organized as follows.
Section 9.4 discusses the use of genetic algorithms for optimizing the CNN templates.
Section 9.5 presents the training algorithm and some simulation results as well. Section 9.6
formulates some concluding remarks. Further, the quintessence of the results obtained is
summarized, and some open research questions are outlined.
