Explore the environment
Avoid collisions
Build the map
Sensors
Actuators
Plan motion
Perceive the environment
Monitor the changes
8
Biologically Inspired Robotics
FIGURE 1.7
Concept of behavior-based control.
Iterative learning control was developed by Arimoto in 1982 based on
observation of the human learning process of drawing lines, circles, and letters. When I was in primary school, whenever the teacher taught us a new
Chinese character, he always asked us to write the character twenty or thirty
times after class. The process of learning new characters is as follows: we
first tried to write the character and then compared what we wrote (real
trajectory in robotics) with the model (desired trajectory) so that we could
correct our hand inputs the next time. By repeating this cycle several times,
we could always write the character well. Iterative learning control actually
simulates this process. Consider the case of a robot drawing a circle using
iterative learning. Denote the desired trajectory of the robot in drawing the
circle by (Td ( )
t ,T q d ( )
t ,T q d ( )
t ) and the real trajectory of the robot at the kth trial
by (T k ( )
t ,T q k (t ),T q k ( )
t ) . Let X k ( )
t represent the joint input of the robot maniplator at the kth trial. \ k ( )
t denotes the output of the robot at the kth trial,
and \ d ( )
t is the desired output. The output could be either the velocity or the
position of the robot. The input X k+1 ( )
t of the robot at the next trial, that is,
the (k + 1)-th trial, is given by the following P-type learning process:
u k +1 ( )
t = u k ( )
t + K (y d ( )
t − y k (t))
(1.1)
where K is the positive-definite learning gain. The asymptotic convergence
of the position error of the robot manipulator under iterative learning control
has been proved using Lyapunov’s theory. Figure 1.8 shows a block diagram
of the iterative learning controller. Arimoto, Naniwa, and Suzuki (1990) also
developed a D-type learning controller and a learning controller with a forgetting factor.
A genetic algorithm is a heuristic optimization method inspired by natural
evolution of biological systems. It uses evolution algorithms to search for the
optimal solution. Genetic algorithms have been used in robotics for optimal
control, planning, etc. Swarm control is inspired by the group behaviors of ants.
A single ant has very limited ability to transport food. However, a group of
Avoid collisions
Build the map
Sensors
Actuators
Plan motion
Perceive the environment
Monitor the changes
8
Biologically Inspired Robotics
FIGURE 1.7
Concept of behavior-based control.
Iterative learning control was developed by Arimoto in 1982 based on
observation of the human learning process of drawing lines, circles, and letters. When I was in primary school, whenever the teacher taught us a new
Chinese character, he always asked us to write the character twenty or thirty
times after class. The process of learning new characters is as follows: we
first tried to write the character and then compared what we wrote (real
trajectory in robotics) with the model (desired trajectory) so that we could
correct our hand inputs the next time. By repeating this cycle several times,
we could always write the character well. Iterative learning control actually
simulates this process. Consider the case of a robot drawing a circle using
iterative learning. Denote the desired trajectory of the robot in drawing the
circle by (Td ( )
t ,T q d ( )
t ,T q d ( )
t ) and the real trajectory of the robot at the kth trial
by (T k ( )
t ,T q k (t ),T q k ( )
t ) . Let X k ( )
t represent the joint input of the robot maniplator at the kth trial. \ k ( )
t denotes the output of the robot at the kth trial,
and \ d ( )
t is the desired output. The output could be either the velocity or the
position of the robot. The input X k+1 ( )
t of the robot at the next trial, that is,
the (k + 1)-th trial, is given by the following P-type learning process:
u k +1 ( )
t = u k ( )
t + K (y d ( )
t − y k (t))
(1.1)
where K is the positive-definite learning gain. The asymptotic convergence
of the position error of the robot manipulator under iterative learning control
has been proved using Lyapunov’s theory. Figure 1.8 shows a block diagram
of the iterative learning controller. Arimoto, Naniwa, and Suzuki (1990) also
developed a D-type learning controller and a learning controller with a forgetting factor.
A genetic algorithm is a heuristic optimization method inspired by natural
evolution of biological systems. It uses evolution algorithms to search for the
optimal solution. Genetic algorithms have been used in robotics for optimal
control, planning, etc. Swarm control is inspired by the group behaviors of ants.
A single ant has very limited ability to transport food. However, a group of