fritzfrancisco / fish_abm

Agent based model for simulating fish movement

fish abm

Agent based model for simulating movement of fish schools.

Image 1: First visualization of the agent based model without environment

Image 2: First visualization of the agent based model with environment

int main():

• create Mat environment & Mat fish

• create_environment()

• individuals fish

• initialize()

  void sample():

  • counting individuals in_zone(r = 15),in_zone(r = 100) and in_zone(r = 200)

    1. when alone lag is decreased [-1] and individual moves to group
    2. if sample_rate > distrib_sample(rd) (random number) lag is set to [5] and individual feeds. This implements that individuals by chance will sample more frequently in a "good" food patch
    3. else, if lag is 0, sample_rate increases [+1]
    4. else, lag is reduced [-1] and sample_rate increases [+1]

  void feed():

  • sample_rate = 10 * q (q being the quality in the quality box encountered)
  • reduce quality of quality box encountered (feeding)

  void move():

  • the direction angle is corrected back between 0 - 360 degrees and the individual moves in this direction

    1. if lag == 0 new direction is calculated dependent on the social force experienced and an added random error of +-10 degrees; Distance traveled depends on the speed_factor (see: double get_speed)Movement can occur across window boundaries.
    2. else, new direction is previous direction with added random error but without social force dependency

  double social():

  • social response is calculated base on the absence or presence of other individuals in_zone(15),in_zone(100) and in_zone(200) and new direction calculated using averageturn()

    1. if the other individuals are within front field of view (330 degrees / 2) and comfort zone (15) focal individual can express:

    2. avoidance (left/right or directly in front) [weight function (1 / ( 0.1 * pow((dist),2) + 2 ))]

    3. if the other individuals are within front field of view (330 degrees / 2) and alignment distance (100) focal individual can express:

    4. alignment to closest neighbor [weight function ((1 / ( 0.004 * pow((dist-15),2) + 2 ))]

    5. if the other individuals are within front field of view (330 degrees / 2) and distance (200) focal individual can express:

    6. attraction to closest neighbor [weight function (1 / ( 0.004 * pow(( dist-200),2) + 2 ))]

    7. alignment to closest neighbor [weight function ((1 / ( 0.004 * pow((dist-15),2) + 2 ))]

  double get_speed():

  • the speed_factor [default=1] is calculated using the individuals in the front visual field (the number of individuals in front (-90,+90 degrees from head tail axis)) and the response radius of 100: speed_factor = 1 + ind_count divided by N/2 (N: Total Population). The more individuals are perceived ahead, the faster the individual will travel.
  

  Image 2: Social force heat-map

  All visualization was done using OpenCV2 v. 2.4.13 in C++.