Protection algorithms using fault resilient fish swarming behaviour

Protection algorithms using fault resilient fish swarming behaviour

Deaconu, Sebastian (TU Delft Electrical Engineering, Mathematics and Computer Science)

Venkatesha Prasad, R.R. (mentor)
Sharma, S. (mentor)
Simha, A. (mentor)
Chen, Lydia Y. (graduation committee)

Delft University of Technology

Computer Science and Engineering

CSE3000 Research Project

2022-06-23

This paper analyzes how flocking behavior in fish can be used to develop target protection algorithms. This starts from the hypothesis that fish aggregate into coordinated flocks in order to protect themselves from predatory attacks. In order to test the protection capabilities of fish, a Prey-Predator instance is developed in which faults are introduced. Prey fish try to protect their faulty flock-mates and themselves from predatory attacks while Predators hunt the fish in order to stay alive. The simulated fish are developed using Boids that emerge in a 70 prey versus 7 predators ecosystem [4]. Results are then extracted using multiple attacking and protection strategies as well as different faulty Boid configurations.
The results initially show no improvement when flocking around faulty prey but when a genetic approach is introduced, the prey gains a clear advantage against the predators. This implies that fish flocking (as opposed to individualistic behavior) is an optimal protection strategy against attacks and could be used in other instances such as military operations or agriculture automation.

Swarm Robotics
Multi-Robot System
Boids
Flocking
Fault resilience
Flock Protection
Distributed Learning

http://resolver.tudelft.nl/uuid:f751b734-6fa9-4ca1-ab5e-940f5682c90f

Student theses

bachelor thesis

© 2022 Sebastian Deaconu