Distributed Broadcast Control of Multi-Agent Systems Using Hierarchical Coordination.

Area of Science:

• Robotics
• Control Systems
• Swarm Intelligence

Background:

• Broadcast control (BC) is a bio-inspired coordination technique for agent swarms.
• A single coordinator broadcasting a signal enables agents to move towards a collective optimal state without direct communication.
• Scaling BC to larger systems with more agents increases convergence time due to limitations of a single coordinator.

Purpose of the Study:

• To propose a revamped broadcast control model using distributed multiple coordinators for efficient control of larger multi-agent systems.
• To enhance the performance and practicality of BC for complex tasks like coverage control.
• To address the convergence time and efficiency challenges of traditional single-coordinator BC models.

Main Methods:

• A hierarchical BC scheme is developed, incorporating distributed sub-coordinators.
• Sub-coordinators broadcast feedback signals received from a global coordinator to agents for task accomplishment.
• A weighted averaging of gradient estimation within a stochastic optimization mechanism is used to manage the dual role of sub-coordinators.

Main Results:

• Numerical simulations for a coverage control task demonstrate the proposed model's effectiveness.
• The hierarchical BC scheme achieves similar convergence to typical BC schemes.
• The proposed scheme shows approximately 30% reduction in traveling costs and reaches near convergence in about one-third of the iteration steps compared to typical BC.

Conclusions:

• The revamped hierarchical BC model with distributed coordinators offers significant improvements in efficiency and convergence speed for large-scale multi-agent systems.
• This approach enhances the practicability of BC for complex coordination tasks.
• The findings highlight the potential of distributed control architectures in swarm robotics and related fields.