A data-driven approach to simulate collective behaviors

Collective behavior is a phenomenon observed in various animal species, where individuals organize as a group, exhibiting coordinated actions. Regarding fish, there is schooling behavior, which has an enormous biological significance and concerns a wide variety of adaptive functions. In this study, we model virtual schooling by using a data-driven approach, more specifically, by using a parameter identification procedure constrained to well-known collective behavioral algorithms. To achieve this, we consider different behaviors such as aggregation, repulsion, target, and leadership mechanisms. Analyzing each fish's behavior inside the schooling it is possible to observe how each behavioral function is set and contributes to global behavior. Then, by incorporating these adjusted algorithms we compare the obtained results with the ground truth. With this, we can create even more realistic simulations replicating the collective behaviors observed in natural fish schools. This approach enables us to understand the collective intelligence of fish schools and harness their adaptive strategies for practical purposes.