• Figure 9-9: Snake Robot rectilinear locomotion Figure 9-8: Wave generated for rectilinear locomotion
  • Figure 9-10: Series 1 is fitness-function value; Series2 is fitness-function minimum value, during cycle of time in learning process. (Series1 is
  • Ultra fast cnn based Hardware Computing Platform Concepts for adas visual Sensors and Evolutionary Mobile Robots




    Download 3,22 Mb.
    Pdf ko'rish
    bet67/81
    Sana16.05.2024
    Hajmi3,22 Mb.
    #238917
    1   ...   63   64   65   66   67   68   69   70   ...   81
    Bog'liq
    Alireza Fasih

    Figure 
    9-9: Snake Robot rectilinear locomotion 
    Figure 
    9-8: Wave generated for rectilinear locomotion 
    05
    .
    3
    ,
    44
    .
    1
    74
    .
    1
    73
    .
    3
    27
    .
    4
    39
    .
    2
    97
    .
    0
    5
    43
    .
    3
    83
    .
    4
    ,
    58
    .
    0
    1
    .
    2
    92
    .
    0
    69
    .
    1
    15
    .
    4
    62
    .
    4
    01
    .
    2
    74
    .
    2
    67
    .
    2
    I
    T
    T
    B
    A


     
    109 
    -5
    0
    5
    10
    15
    20
    25
    30
    35
    1
    67 133 199 265 331 397 463 529 595 661 727 793 859
    Series1
    Series2
    Figure 
    9-10: Series 1 is fitness-function value; Series2 is fitness-function 
    minimum value, during cycle of time in learning process. (Series1 is 
    error rate; Series2 is number of itteration/time) 
    Figure 
    9-11: Learning 4-legs semi-spider 
    robot 
    Iterations 
    Erro

    P
    er
    centage


     
    110 
    Figure 9-11 shows a spider robot with 4 legs and 16 degrees of freedom. Each hinge has 
    2 degrees of freedom in rotation. A 16×16 CNN array can be used to drive this robot. Figure 
    9-12 shows the sequences of the robot locomotion after the learning process. In this test, 
    the robot must turn around the ‘z’ axis. Another test in Figure 9
    -11 shows the design of a 
    6 legs insect robot for locomotion learning. The robot has 12 degrees of freedom in hinges. 
    In this case, the aim is moving around the circle with a given radius. After nearly 2500 
    iterations it was found that the result converged to zero. The CNN templates shown in 
    Equation 9-6 are optimized for this purpose. The fitness function in Equation 9-7 is used to 
    generate the CNN output wave shown in Figure 9-14. 
    (9-6)

    Download 3,22 Mb.
    1   ...   63   64   65   66   67   68   69   70   ...   81




    Download 3,22 Mb.
    Pdf ko'rish

    Bosh sahifa
    Aloqalar

        Bosh sahifa



    Ultra fast cnn based Hardware Computing Platform Concepts for adas visual Sensors and Evolutionary Mobile Robots

    Download 3,22 Mb.
    Pdf ko'rish