Progressive organization of co-operating colonies/collections of ants/agents (POOCA) for competent pheromone-based navigation and multi-agent learning
Book chapter, 2011
Progressive Organization of co-Operating Colonies/Collections of Ants/Agents (POOCA) is introduced as a novel robot navigation and obstacle avoidance approach that is capable of efficiently as well as accurately operating in both stationary and dynamic unknown environments. POOCA follows the streak of existing pheromone-based co-operative foraging multi-agent models in its biological inspiration (colonies of ants foraging their environment in search of food) and the utilization of evolutionary computation for optimizing the various parameters involved, but differs from them in its simulation of minimal and uniformly applied ant-inspired notions. POOCA's main principles include: stigmetry; stereo-like sensors and a single type of pheromone; inertia to change in direction of motion; intensifying pheromone laying, constant-rate pheromone evaporation, diminishing pheromone diffusion. In addition, near-minimum ant/agent colony/collection sizes that allow speedy convergence to (near-) optimal trails are determined. Simulations demonstrate POOCA's potential as a competent robot navigation and obstacle avoidance approach that imitates minimal biological principles, while also putting it forward as a useful emerging paradigm in swarm robotics, ant colony optimization and co-operative multi-agent learning.
Co-operative multi-agent learning
Optimal trail creation.
Ant colony optimization