Partical Systems: Animating Fish Behaviors

Rick Shanor and Pratik Chatrath

For this assignment, we developed a 2D fish-tank simulator to animate schooling/shoaling behavior of fishes based on particle interactions. Our final demo shows an animation consisting of prey (schools of fishes) and predators (sharks) that is forward simulated based on simple interactions between the agents in the environment. Prey have position, velocity and heading attributes. Attraction and repulsion forces generate swarm behaviors. Predators have similar attributes. Predators have the ability to eat prey. Additionally, mouse click on a particular position scares nearby fish. A motivating image is shown below. Most of our code is in the files fish.cpp and fish.h. We also extended the viewer to support animation.

### User Guide - How to run the animation

First build the project

mkdir build && cd build && cmake .. && make

Next, run the program from the command line

./aquarium #fish #turtles #sharks

Finally, press spacebar to start and stop the animation. Click with the mouse to scare fish.

Demo Videos

Video demonstrating the simulation initialized with: 50 Minnow fishes, 1 turtle and 4 sharks. Fishes in the same group have same color [![Alt text] (https://img.youtube.com/vi/-whSFRU2zZg/maxresdefault.jpg)] (https://youtu.be/-whSFRU2zZg)

Links to more videos:

Scare Feature - (https://youtu.be/imRcG70os7w)

100 fishes - (https://youtu.be/7FLz5lNjPaA)

Implementation Details

The following sequence of steps were proposed for implementation of the project:

1. Create/find SVGs for fish and sharks and extend HW1 rasterizer as necessary to draw these files in the viewer.
2. Create classes for fish and sharks. Pull in relevant animation code from HW4 such that we can forward simulate objects motion based on position and velocity.
3. Write code calculating the repulsive and attractive forces between agents in the environment. Using a K-D tree structure may be required to search efficiently.
4. Integrate forces calculated in #3 to update the velocities in #2. At this point, we should have the minimally viable particle simulator in place.
5. Extend prey class for reproduction and predator class for feeding to generate interesting, natural behaviors.

The following sequence of steps were actually followed during the implementation of this project:

1. SVG files for fish and sharks were generated from PNGs. Our svg parser supports svg files with polygon points list. The generated svg files are defined in terms of polygon and are supported by our SVGParser
2. Created class for fish. Each fish has its position, velocity and heading. Extended Fish class to Minnow (kind of fish) and shark classes. Updated DrawSVG time_event to update position, velocity, heading at each instance of time.
3. To calculate the updated position, velocity, heading of each fish at a time instance following rules were implemented
   • Fish of same kind attract each other
   • If fishes of same type come very close to each other they experience repulsive forces
   • Minnow fish tries to avoid shark
4. Using rules mentioned in step 3 a functioning school of fish behavior was obtained
5. Sharks consume Minnow fishes which are very close to it

Extra features implemented:

1. Mouse clicks exhibit a fish scare behavior
2. Added support for displaying turtle
3. Fish of the same group have same color (average color of all the fishes in group)

Difficulties in implementation

1. We faced issue bounding fish movement to a confined fish tank. We got rid of this issue by creating unbounded fish tank. The complete screen is now used as fish tank and fish can enter/exit from any direction
2. Secondly, making Minnow fish avoid sharks wasnt easy. The minnow fishes passed through sharks. We worked around this issue by creating a bounding box around the shark svg and making the fish avoid the bounding box.
3. The most difficult part was to implement natural/realistic fish behavior. We worked on fine tuning the attraction, repulsion, and other constants related to updating position, velocity, heading to get realistic behavior.

Learning outcomes of this assignment

1. Better understanding of hardware renderer, svg parser, animation
2. First hand experience implementing mobile robot kinematics & simulating particle interactions

References

This assignment was completed for Carnegie Mellon's Computer Graphics class (15-462), taught by Professors Keenan Crane and Stelian Coros. More info can be found on the course website here: http://15462.courses.cs.cmu.edu/fall2016/home

[1] School of fish simulation video: https://www.youtube.com/watch?v=NKhHp7Wfdmk

[2] Schooling fish behavior: https://www.mrl.ucsb.edu/sites/default/files/mrl_docs/ret_attachments/research/KTaylor.pdf

[3] Image: https://s-media-cache-ak0.pinimg.com/564x/b4/87/a1/b487a19e031b7d220e5638cd5c69c109.jpg

[4] Flocking video: https://www.youtube.com/watch?v=QbUPfMXXQIY&feature=youtu.be

[5] Coordinated motion: http://www.red3d.com/cwr/boids/