Collective evolution learning model for vision-based collective motion with collision avoidance.

Area of Science:

• Computational neuroscience
• Artificial intelligence
• Behavioral ecology

Background:

• Collective motion (CM) and collision avoidance (CA) are fundamental behaviors observed across various natural systems, including fish schools, bird flocks, and insect swarms.
• These behaviors are typically driven by environmental inputs like vision, smell, and air pressure, processed individually to guide group dynamics.
• Understanding the learning mechanisms underlying emergent CM and CA is crucial for modeling complex biological systems.

Purpose of the Study:

• To propose a novel vision-based collective motion with collision avoidance (VCMCA) model.
• To simulate the collective evolution learning process in multi-agent systems.
• To investigate the emergence of global CM and CA dynamics from individual learning processes.

Main Methods:

• Development of a vision-based computational model (VCMCA) where learning agents perceive their environment visually.
• Implementation of a trial-and-error learning approach for individual agents to acquire local CM and CA behaviors.
• Evaluation of the VCMCA model using simulations of locust swarms to observe the evolution of collective behaviors.

Main Results:

• The VCMCA model successfully simulates the emergence of local CM and CA behaviors through individual learning.
• The model demonstrates how these learned individual behaviors scale up to produce global CM and CA dynamics within a swarm.
• Simulations using locust swarms validated the biologically-inspired learning process and its ability to generate complex multi-agent dynamics.

Conclusions:

• A biologically-inspired learning process can effectively generate multi-agent, multi-objective dynamics for collective motion and collision avoidance.
• Vision-based learning is a viable mechanism for individuals to develop complex collective behaviors.
• The VCMCA model provides a novel framework for studying the evolution of collective behaviors in natural systems.